WO R K I N G PA P E R S E R I E S
N O 170 4 / A U G U S T 2014
ARE FOREIGN-OWNED
FIRMS DIFFERENT?
COMPARISON OF EMPLOYMENT
VOLATILITY AND ELASTICITY
OF LABOUR DEMAND
Jaanika Meriküll and Tairi Rõõm
THE COMPETITIVENESS
RESEARCH NETWORK
In 2014 all ECB
publications
feature a motif
taken from
the €20 banknote.
NOTE: This Working Paper should not be reported as representing
the views of the European Central Bank (ECB). The views expressed are
those of the authors and do not necessarily reflect those of the ECB.
CompNet
The Competitiveness Research Network
T
This paper presents research conducted within the Competitiveness Research Network (CompNet). The network is composed of
economists from the European System of Central Banks (ESCB) - i.e. the 28 national central banks of the European Union (EU) and the
European Central Bank – a number of international organisations (World Bank, OECD, EU Commission) universities and think-tanks,
as well as a number of non-European Central Banks (Argentina and Peru) and organisations (US International Trade Commission).
The objective of CompNet is to develop a more consistent analytical framework for assessing competitiveness, one which allows for a
better correspondence between determinants and outcomes.
The research is carried out in three workstreams: 1) Aggregate Measures of Competitiveness; 2) Firm Level; 3) Global Value Chains
CompNet is chaired by Filippo di Mauro (ECB). Workstream 1 is headed by Chiara Osbat, Giovanni Lombardo (both ECB) and
Konstantins Benkovskis (Bank of Latvia); workstream 2 by Antoine Berthou (Banque de France) and Paloma Lopez-Garcia (ECB);
workstream 3 by João Amador (Banco de Portugal) and Frauke Skudelny (ECB). Julia Fritz (ECB) is responsible for the CompNet
Secretariat.
The refereeing process of CompNet papers is coordinated by a team composed of Filippo di Mauro (ECB), Konstantins Benkovskis
(Bank of Latvia), João Amador (Banco de Portugal), Vincent Vicard (Banque de France) and Martina Lawless (Central Bank of Ireland).
The paper is released in order to make the research of CompNet generally available, in preliminary form, to encourage comments and
suggestions prior to final publication. The views expressed in the paper are the ones of the author(s) and do not necessarily reflect those
of the ECB, the ESCB, and of other organisations associated with the Network.
Acknowledgements
The authors thank discussants of the paper at the following meetings: the annual conference of the Estonian Association of Economists
in 2013; the NOeG conference in Innsbruck; the SMYE conference in Aarhus; the EACES workshop in Tartu; and the CompNet
meeting in Frankfurt. Jaanika Meriküll is grateful for the financial support from the European Social Fund and Estonian Science
Foundation grant no. 8311. The views expressed are those of the authors and do not necessarily represent
the official views of Eesti Pank.
Jaanika Meriküll
Eesti Pank, University of Tartu; e-mail: [email protected]; [email protected]
Tairi Rõõm
Eesti Pank, Tallinn University of Technology; e-mail: [email protected]
© European Central Bank, 2014
Address
Postal address
Telephone
Internet
Kaiserstrasse 29, 60311 Frankfurt am Main, Germany
Postfach 16 03 19, 60066 Frankfurt am Main, Germany
+49 69 1344 0
http://www.ecb.europa.eu
All rights reserved. Any reproduction, publication and reprint in the form of a different publication, whether printed or produced
electronically, in whole or in part, is permitted only with the explicit written authorisation of the ECB or the authors. This paper can
be downloaded without charge from http://www.ecb.europa.eu or from the Social Science Research Network electronic library at
http://ssrn.com/abstract_id=2464134. Information on all of the papers published in the ECB Working Paper Series can be found on the
ECB’s website, http://www.ecb.europa.eu/pub/scientific/wps/date/html/index.en.html
ISSN
ISBN
EU Catalogue No
1725-2806 (online)
978-92-899-1112-2 (online)
QB-AR-14-078-EN-N (online)
Abstract
This paper analyses differences in employment volatility in
foreign-owned and domestic companies using firm-level data
from 24 European countries. The presence of foreign-owned
companies may lead to higher employment volatility because
subsidiaries of multinational companies react more sensitively to
changes in labour demand in host countries or because they are
more exposed to external shocks. We assess the conditional employment volatility of firms with foreign and domestic owners using propensity score matching and find that it is higher in foreignowned firms in about half of the countries that our study covers.
In addition, we explore how and why labour demand elasticity
differs between these two groups of companies. Our estimations
indicate that labour demand can be either more or less elastic in
subsidiaries of foreign-owned multinationals than in domestic enterprises, depending on the institutional environments of their
home and host countries.
JEL Codes: F23, J23, J51
Keywords: foreign direct investment (FDI), employment volatility, labour
demand, labour market institutions, European Union
ECB Working Paper 1704, August 2014
1
Non-technical summary
There is a long-running debate about the potential adverse side effects of
globalisation. The increase in employment volatility is one of the side effects
usually depicted in a negative light, since it lessens job security (Scheve and
Slaughter (2004); Geishecker et al. (2012)). Globalisation could increase the
volatility of employment for two main reasons. First, internationalization of
production may amplify the volatility of shocks that firms face (Bhagwati
(1996)). Second, multinational companies may react to shocks more strongly,
i.e. their elasticity of labour demand could be larger in absolute value.
An explanation why the second of the above-stated reasons can be the
cause of increasing employment volatility was given by Rodrik (1997). He
alleged that deeper international economic integration would make domestic
workers more easily substitutable by foreign workers – the elasticity of substitution of labour would increase. Consequently, labour demand would become more wage (or own-price) elastic. This result of globalization has been
debated by Hijzen and Swaim (2010) who argue that the effect of globalization on the elasticity of labour demand is theoretically ambiguous and can be
only empirically determined.
This paper analyses how one aspect of globalization – the internationalization of the ownership structure – contributes to employment volatility. Using
the standard framework of labour demand and supply, we show that the differences in total employment volatility can be caused either by the foreignowned firms’ different elasticity of labour demand or by their different exposure to economic shocks. The empirical analysis focuses on two aims. First,
we evaluate differences in employment volatility in foreign-owned and domestically owned enterprises (FOEs and DOEs). Second, we assess whether
these differences stem from the elasticity of labour demand. We employ firmlevel panel data from the Bureau van Dijk’s Amadeus database, which covers
24 European countries and spans the years 2001–2009. We assess the conditional employment volatilities of FOEs and DOEs using propensity score
matching, which enables us to control for differences in firm characteristics
such as age, size, capital intensity, labour productivity, ownership concentration, number of subsidiaries and primary business activity. A comparison of
conditional employment volatilities implies that FOEs tend to have higher
employment volatility than DOEs with similar characteristics. This difference
is statistically significant in about half of the countries that our study covers.
The average magnitude of the difference across all covered countries is
around 10%, i.e. FOEs have approximately 10% higher employment volatility.
Besides assessing the volatility of employment, we estimate labour demand equations for FOEs and DOEs using system GMM estimations and
ECB Working Paper 1704, August 2014
2
find that there are only a few European countries where labour demand elasticities of the two groups differ to a statistically significant degree. There is
no conclusive result that elasticity of labour demand is higher in FOEs. The
results are country-specific, indicating for example that foreign-owned firms
have more elastic labour demand in Italy and Belgium and less elastic labour
demand in France and Spain.
Given these findings, our study implies that although employment volatility tends to be higher in FOEs than in DOEs, this gap in volatility is not unanimously caused by their more elastic labour demand. We analyse the determinants of the elasticity of labour demand in foreign-owned multinationals
further by assessing the role that labour market institutions play in this context.
Our estimations indicate that labour demand can be either more or less
elastic in the subsidiaries of foreign-owned multinationals than in DOEs,
depending on the institutional environments of their home and host countries.
When FDI originates from a region with a more flexible institutional environment (e.g. from the USA to Western European countries) then the elasticity of labour demand is smaller in absolute value in FOEs than in DOEs. In
the opposite case (e.g. when FDI is originating from Germany to CEE countries) the elasticity of labour demand is higher.
A potential explanation for this finding is that in countries with rigid labour market regulations, multinational companies avoid changing domestic
employment in response to economic shocks and instead use other margins of
adjustment. Alternatively, multinational firms may choose the host countries
where they establish subsidiaries by looking at the labour market institutions:
if they operate in sectors that have highly volatile demand then they are more
likely to move to countries with a flexible institutional environment and vice
versa. In either case, the presence of the subsidiaries of foreign-owned multinationals would have an amplifying effect on the elasticity of labour demand
in countries with flexible labour market institutions, whereas it would have a
dampening effect in countries with rigid institutions.
ECB Working Paper 1704, August 2014
3
1. Introduction
There is a long-running debate about the potential adverse side effects of
the internationalisation of ownership structures and those of globalisation in
general. The increase in employment volatility is one of the side effects usually depicted in a negative light, since it lessens job security (see e.g. Scheve
and Slaughter (2004) and Geishecker et al. (2012)).1 We study differences in
employment volatility between firms with domestic and foreign workers in
Europe. For this purpose, we use firm-level panel data from Bureau van Dijk
Amadeus database spanning the years 2001–2009. The Amadeus dataset includes a detailed description of firms’ ownership structure, which enables us
to disentangle companies by ownership type and to identify the number of
subsidiaries for multinational and domestic enterprises.
Rodrik (1997) in his book “Has globalization gone too far?” is seen as the
first to argue forcefully that the labour demand of foreign-owned companies
is more elastic, contributing to higher employment volatility and lower job
security. He alleges that deeper international economic integration may make
domestic workers more easily substitutable by foreign workers. Consequently, labour demand would become more wage (or own-price) elastic.
Another reason why globalisation increases the elasticity of labour demand is that deepening international integration of production results in more
elastic product demand. This is an often-cited finding from the empirical literature on international trade and FDI flows. According to the HicksMarshall laws of derived demand, more competition in the product markets
(i.e. flatter product demand curves) should also lead to more elastic labour
demand. Bhagwati (1996) stressed a related channel through which globalisation may have increased employment volatility when he pointed out that
global economic integration has made product markets more volatile. Greater
volatility of product demand should lead to greater volatility of labour demand as well, since the latter is derived from the former.
An alternative view of the relationship between the international integration of production and the elasticity of labour demand is proposed by Hijzen
and Swaim (2010). They argue that the impact of FDI on the elasticity of
labour demand is theoretically ambiguous and hence ultimately an empirical
issue. While the internationalisation of the production process is expected to
increase the ability of firms to substitute between factor inputs, the elasticity
of substitution is only one of several factors determining the own-price elasticity of labour demand. Globalisation, which is associated with greater capi1
The other effects of globalisation remain beyond the scope of this paper. In particular,
the paper does not seek to undermine the positive effects of FDI (see e.g. Borensztein et al.
(1998) on FDI and growth).
ECB Working Paper 1704, August 2014
4
tal mobility, will also tend to lead to a reduction in the cost share of labour.
Making use of a decomposition of the determinants of labour demand elasticity into substitution and scale effects along the lines of Hamermesh (1993),
Hijzen and Swaim (2010) demonstrate that a simultaneous increase in the
constant-output elasticity of substitution and a decrease in the cost share of
labour in production will have offsetting effects on the total own-price elasticity of labour demand. The former will increase elasticity via the substitution effect, while the latter will decrease it via the scale effect. The result is
that the net impact of globalisation can be either positive or negative, depending on which of the two effects dominates.
Given the arguments outlined above, it is not a priori clear that a positive
association exists between foreign ownership and employment volatility. The
empirical evidence is mostly in favour of the existence of this relationship,
but not universally so. Some examples in favour are studies by Bergin et al.
(2009) and Levasseur (2010), which compare employment volatilities in specific offshoring industries in home and host countries. In Bergin et al.’s paper, the country pair is the USA and Mexico, and in Levasseur’s study, Germany is compared with the Czech Republic and Slovakia. Both of these articles focus on specific industries where the vertical integration of production
is well documented and yield the result that employment is more volatile in
the host country in an industry that specialises in subcontracting.
However, studies analysing a wider spectrum of industries and incorporating services in addition to manufacturing do not always yield the result that
globalisation is associated with increasing labour volatility. For example, an
analysis by Buch and Schlotter (2013) using German industry-level data
demonstrates that unconditional volatility of employment has exhibited a
downward trend. According to this study, openness to trade and employment
volatility are not significantly related across industries in Germany.
Most of the research papers investigating the labour market impacts of
offshoring (or FDI more particularly) focus on the elasticity of labour demand. As explained above, the flattening of the demand curve is one factor
that can contribute to an increase in employment volatility. The results of
these studies are inconclusive. The evidence in support of the hypothesis that
an increase in offshoring leads to more elastic labour demand is provided by
several studies.2 On the other hand, research which has used data from various European countries mostly does not support this hypothesis.3 Among
studies using plant-level or firm-level data, the only case where the higher
2
Supporting evidence can be found in Slaughter (2001) on the US data; Fabbri et al.
(2003) for the UK; and Görg et al. (2009) for Ireland.
3
Examples include Barba Navaretti et al. (2003); Buch and Lipponer (2010); and Hakkala et al. (2010).
ECB Working Paper 1704, August 2014
5
labour demand elasticity of foreign multinationals has found empirical support is in Ireland (Görg et al. (2009)).
The purpose of our study is to assess the differences in employment volatility between firms with domestic and foreign owners. Using the standard
framework of labour demand and supply, we show that the differences in
total employment volatility can be caused either by the foreign-owned firms’
different elasticity of labour demand or by their different exposure to economic shocks. We assess the conditional employment volatilities of firms
with foreign and domestic owners using propensity score matching, which
enables us to control for differences in firm characteristics such as age, size,
capital intensity, labour productivity, ownership concentration, and number
of subsidiaries. A comparison of conditional employment volatilities implies
that foreign-owned firms tend to have systematically higher employment
volatility than domestically owned counterparts with similar characteristics,
although this difference is not statistically significant for all the countries that
our study covers.
Regarding the elasticity of labour demand, we do not find evidence to
support Rodrik’s (1997) conjecture described above. The system GMM estimations of labour demand functions across 18 European countries indicate
that the wage elasticity of labour demand is mostly not significantly different
between foreign and domestically owned enterprises. For the few countries
where the differences are significant the elasticity is not always larger in foreign-owned firms. The main focus of our analysis is on assessing the role that
labour market institutions play in this context.
The results of two earlier studies indicate that the effect of offshoring or
foreign ownership on the elasticity of labour demand is dependent on labour
market institutions. Barba Navaretti et al. (2003) show that long-term wage
elasticity of labour demand is lower in multinational enterprises (MNEs) than
in domestic firms and the ratio of the elasticities of MNEs and NEs is larger
in countries with a stricter institutional environment. They argue that MNEs
manage to bypass the regulations in a strict regulatory environment and conclude that “labour market regulations are quite irrelevant to the labour market
behaviour of MNEs” (Barba Navaretti et al. (2003, p. 718). The analysis of
Hijzen and Swaim (2010) indicates that offshoring is associated with higher
labour demand elasticity only in countries with relatively weak employment
protection legislation, whereas they detect no significant effects for countries
with more regulated labour markets.
In comparison to the earlier research, we take a step further and investigate the role of labour market institutions in a bilateral context by assessing
the effects of differences in the institutional environment in the home and
host countries of MNEs. We find that labour demand can be either more or
ECB Working Paper 1704, August 2014
6
less elastic in subsidiaries of foreign-owned multinationals than in domestic
enterprises, depending on these institutional differences. When FDI originates from a region with more flexible institutions then the elasticity of labour demand is smaller in absolute value in foreign-owned firms. In the opposite case the elasticity of labour demand is higher. A potential explanation
for this empirical finding is that it is easier for multinational companies to
substitute between factor inputs and so they have more flexibility than domestic firms in choosing which channels of adjustment to use.
When MNEs need to adjust costs in response to economic shocks, then in
the presence of strong restrictions on the adjustment of employment it is easier for them to alter other production costs or output prices and leave labour
costs unadjusted. A multinational production network should be associated
with easier adjustment via other margins than is the case for companies that
have only domestic operations. In addition, MNEs can respond to shocks by
adjusting employment in other locations abroad. If it is necessary to change
employment in response to economic shocks then they can shift adjustments
to countries or regions where it is easier to adjust. They can change employment mostly at home when the labour market there is more flexible or shift
the main bulk of adjustment to foreign affiliates when the local institutions in
the host countries favour this.
It is worth noting that we use a similar explanation for our empirical findings to that evoked by Rodrik (1997). He asserted that multinational enterprises have larger elasticity of substitution between production factors and
this should increase their elasticity of labour demand. We add another layer
to this argument as our empirical estimates imply that this greater ease of
substituting between different inputs can also result in smaller elasticity of
labour demand, depending on labour market institutions. Differences in institutional environment can lead to a dual outcome: the presence of MNEs can
have an amplifying effect on the elasticity of labour demand in countries with
flexible labour market institutions, whereas it can have a dampening effect in
countries with rigid institutions.
An alternative, though related, explanation for this empirical finding is
that multinational firms choose the host countries where they will establish
subsidiaries by looking at the labour market institutions: if MNEs operate in
sectors characterised by highly volatile demand then they are more likely to
move to countries with a flexible institutional environment. The formalisation
of how flexible labour markets act as a comparative advantage is provided
e.g. in Cunat and Melitz (2012).
The paper is organised as follows. The second section presents the theoretical model deriving the decomposition of employment volatility. The third
section provides an overview of the Bureau van Dijk Amadeus firm-level
ECB Working Paper 1704, August 2014
7
data that we employ for the analysis. In the fourth section, we give an overview of unconditional and conditional employment volatilities for foreign
and domestically owned firms. Section 5 focuses on estimating labour demand equations for foreign and domestically owned firms and investigating
the role of labour market institutions. The last section summarises.
2. Decomposition of employment volatility
The subsidiaries of foreign-owned enterprises can have higher volatility
than local companies for two reasons. First, they may be exposed to more
volatile shocks, which can then be transferred into more volatile labour demand, and second, they may behave differently from local enterprises as they
can react to shocks of similar size more or less strongly by adjusting labour.
This section will derive a decomposition of employment volatility into two
subcomponents: a) a function of exogenous economic shocks; and b) a function of the elasticities of labour supply and demand. This decomposition will
enable us to demonstrate that employment volatility is positively related to
the elasticity of labour demand as long as labour supply is not perfectly inelastic. This can be assumed to be the case if the subject of the analysis is a
firm, as in the current study.
We build on the approach of Scheve and Slaughter (2004) and Barba
Navaretti and Venables (2004) along the lines of Hamermesh (1993) to decompose employment volatility. Let us assume a Cobb-Douglas production
function with diminishing returns to scale where capital is fixed in the shortterm and normalised to one:
(1)
where Y denotes output, A is the parameter capturing technological progress
and L denotes labour, while 0 < β < 1. Profit maximisation under perfect
competition in all markets yields:
(2)
where W stands for wages, p is product price and the term pAβ is marginal
revenue product, which captures exogenous price and productivity shocks.
Solving for L and defining labour demand as LD results in the following labour demand equation:
/
ECB Working Paper 1704, August 2014
(3)
8
Given that the labour demand elasticity equals 1 / (β–1) in this case and
defining ηLL as the absolute value of the wage elasticity of labour demand lets
us rewrite equation (3) as:
(3’)
Let us assume the following labour supply function:
,
(4)
where ηS denotes the wage elasticity of labour supply. The equilibrium employment and wage can then be expressed as follows:
/
(5)
/
(6)
Taking natural logarithms of both sides of equations (5) and (6) (a monotonic transformation) yields:
/
ln
(7)
/
ln
(8)
where w = ln(W) and l = ln(L).
Treating marginal revenue product as a random variable, we can express
the variance of equilibrium employment and wages by building on equations
(7) and (8) as follows:
/
/
(9)
(10)
Equation (9) implies that employment volatility can be expressed as a
combination of two components. The first part, in square brackets, captures
volatility in employment due to changes in labour demand elasticity. Given
non-zero finite elasticity of labour supply, the elasticity of labour demand is
positively related to employment volatility, ceteris paribus. The second part
captures volatility in employment due to changes in the exposure to economic shocks. The more exposed a firm is to external shocks or the higher the
ECB Working Paper 1704, August 2014
9
variation in marginal revenue product is, the higher its employment volatility
is.
Note that when the labour supply is perfectly inelastic then changes in the
elasticity of labour demand do not affect employment volatility. On the other
hand, equation (10) implies that when the labour supply is perfectly elastic
then changes in the elasticity of labour demand do not affect wage volatility.
In general, the distribution of volatility between wages and employment depends on the slope of the labour supply curve. The more elastic it is, the larger employment volatility is relative to wage volatility, given a similar demand
schedule and exogenous shocks to labour demand. Since labour market rigidities make the labour supply less elastic, it can be expected that employment
will be more volatile in countries with flexible labour regulations, ceteris
paribus.
The decomposition given in equation (9) illustrates that foreign-owned
companies may have higher employment volatility because they react more
sensitively to wage changes in a host country or because they are more exposed to external shocks. The latter might well be the case since foreignowned MNEs are more likely to operate in several markets and to be hit by
shocks more frequently than domestically owned enterprises.4 However, multinationals may also be faced by a more dispersed structure of shocks, so
whether they are more or less exposed to a volatile economic environment is
an empirical issue that depends on the cross-country correlation of shocks.
3. The data
We use an Amadeus (Bureau van Dijk, see https://amadeus.bvdinfo.com)
firm-level panel dataset that covers a large set of European countries and
spans the years 2001–2009. Amadeus data includes information about the
balance sheets and profit/loss statements of firms and detailed information on
the ownership structure.
Our initial goal was to cover all the EU27 countries, but the set of countries was reduced to 18 because of data availability. The Amadeus data on
Greece and Lithuania do not cover employment costs while the data on Ireland do not cover employment volumes. The Amadeus data on Austria, Cyprus, Denmark, Hungary, Latvia, Luxembourg and Malta do not have enough
4
The focus in the current study is on comparing foreign and domestically owned companies. Practically all of the former are subsidiaries or affiliates of multinational companies.
Although some of the domestically owned firms are also multinationals, the majority of
firms in this group are local companies. Thus, as a group, foreign-owned firms can be expected to be more exposed to shocks.
ECB Working Paper 1704, August 2014
10
observations to be suitable for econometric analysis. Our analysis includes
Norway in addition to the EU member states. The default dataset covers 18
countries, 170 thousand firms and in total more than a million observations.
In some cases, like when data on wage costs is not necessary for the analysis,
the set of countries covered is larger. The variables for the empirical analysis
are defined in Table 1.
Table 1: Variable definitions
Variable
Employment (empl)
Wage (rwage)
Definition
Number of employees, head counts
GDP deflator* deflated employment costs divided by employment
Output (rturn)
GDP deflator* deflated turnover (operational revenue for Denmark, Norway, UK)
Foreign-owned enForeign versus domestically owned enterprises (FOEs; DOEs),
terprise (FOE)
dummy variable. A firm is considered to be foreign-owned if
its global ultimate owner is a foreigner (subsidiary) or its largest shareholder is a foreigner (associate). Ownership is timeinvariant and fixed in the year 2009.
Age
Firm’s age in years
No of subsidiaries
Number of recorded subsidiaries
No of shareholders
Number of recorded shareholders
Peer’s employment
Employment of the business group or the largest recorded
owner
Capital intensity
Total fixed assets per employee in real terms
Labour productivity Deflated turnover divided by employment
Notes: The GDP deflator is taken from Eurostat and is at a 2-digit NACE 2008 level.
The ownership data are often missing in the Amadeus dataset. For some
countries like Romania and Slovakia the data are only available for a small
number of companies. The number of observations across the dynamic dimension of the dataset is smaller than average for Germany as the years
2007–2009 are missing for almost all the firms. In general, larger firms tend
to be overrepresented in the Amadeus sample in comparison to the whole
population of firms.
We also impose filters to remove possibly erroneous observations and
make the dataset more comparable across countries. These filters differ for
matching and dynamic panel data analysis and these differences are discussed
in the sections that cover these topics. Country-by-country estimations use
ECB Working Paper 1704, August 2014
11
monetary variables in their original currency, while estimations with pooled
data across countries employ monetary variables transformed into euros5.
Appendix 1 presents the descriptive statistics of variables for foreign and
domestically owned enterprises (FOEs and DOEs) separately for countries
from Western Europe and from Central and Eastern Europe. The foreignowned firms tend to be larger, to pay higher wages, to have higher capital
intensity and labour productivity, to have more concentrated ownership and
to operate more often in the manufacturing sector. In total, 18% of firms are
foreign-owned in the final sample, while 30% of employment originates from
foreign-owned companies. The sample of enterprises from Western Europe
contains some very large firms, which make the samples of WE and CEE
differ much more in the mean values of the variables analysed than in the
medians.
Figure 1 presents the origin of foreign investment from the host country
perspective. FDI in EU countries mostly originates from other EU countries
and is highly concentrated in terms of origins, with Germany, France, the
Netherlands and the UK being the main home countries. Outside the EU the
main country of origin is the USA. Central and Eastern Europe is an important recipient of FDI from Western Europe but the FDI flows from Central
and Eastern Europe to other EU countries are modest.
Our dataset imposes some limitations on what we can or cannot test. First,
we cannot observe firm entry and exit in our data, which means that we can
investigate firms’ employment adjustment only via the intensive margin. Second, we do not cover employment across different skill groups as we only
have data on total wages and employment. Third, our database consists of the
balance sheets and profit/loss statements on a yearly basis but only includes
ownership data for the year 2009, so it is possible that the firm ownership
variable is subject to measurement error.
5
The source of the exchange rates is the European Central Bank Statistical Data Warehouse:
annual
average
bilateral
exchange
rates.
[http://sdw.ecb.europa.eu/browse.do?node=2018794]
ECB Working Paper 1704, August 2014
12
Home country
other
other EU
CH
JP
US
SI
SK
RO
PL
EE
CZ
BG
GB
SE
ES
PT
NO
NL
IT
DE
FR
FI
BE
BE FI FR DE IT NL NO PT ES SE GB BG CZ EE PL RO SK SI
Host country
Figure 1: Country of origin of foreign enterprises (2005)
Notes: Foreign ownership is weighted by employment. See International Standard Codes for the Representation of the Names of Countries (version
2002) for the country abbreviations.
Source: Authors’ calculations from the Amadeus dataset.
ECB Working Paper 1704, August 2014
13
Trade and foreign ownership are sometimes difficult to disentangle. For
example, part of production can be outsourced abroad to another company or
a subsidiary can be established abroad to do this work within a business
group. Offshoring is usually defined as a change in the supplier of intermediate inputs and services from a domestic one to a foreign one. Offshoring can
be international outsourcing, which means importing goods from other firms,
or it can be the relocation of a firm’s own production so that some parts of
the value-added chain are produced abroad within an affiliate or subsidiary.
This relocation is also called in-house offshoring. OECD (2007) notes that
offshoring via the establishment of a new affiliate is more common when
OECD countries are offshoring to other developed countries. When OECD
countries offshore to less developed countries the most common type of offshoring is usually subcontracting. Most of the host countries covered in this
study are OECD countries, meaning that in-house offshoring should be the
most common type of offshoring to these countries and this is what our database captures.
4. Unconditional and conditional employment volatility
In this section we will look at employment volatility across 24 European
countries6, differentiating between foreign and domestically owned enterprises. We start out by comparing the unconditional employment volatilities of
FOEs and DOEs. This comparison performs a simple test as to whether firmlevel employment volatility differs for these two firm groups, i.e. whether the
overall volatility differs in the left-hand side of equation (9). Volatility is
measured as a coefficient of variation (CV) for the time period 2001–2009.
For better comparability, firms with fewer than 5 observations are excluded.
Next, to account for firm heterogeneity, we estimate conditional employment volatilities. We use propensity score matching with the nearest neighbour and a caliper (maximum propensity score distance) algorithm. As it is
sometimes difficult to find a common support for treatment and artificial
counterfactual groups, we match the three nearest neighbours and introduce a
caliper of 0.05 or 0.10, meaning the three nearest neighbours are selected
within a propensity score of 5% or 10%. A caliper of 10% is used in countryby-country analysis, and a caliper of 5% in the analysis of country groups.
We use matching with replacement, meaning that the same firms from the
artificial counterfactual can be used more than once as a match. (See Calien6
We were able to increase the set of countries analysed here by adding Austria, Denmark, Greece, Hungary, Latvia and Lithuania as the employment and ownership data for
these countries was available for a substantial number of firms, unlike the wage costs needed
for the forthcoming sections.
ECB Working Paper 1704, August 2014
14
do and Kopeinig (2008) for a discussion of options for matching algorithms
and Leuven and Sianesi (2003) for psmatch2 module for Stata).
We use control variables from 2005 and estimate the conditional volatility
as a cross-section over this period of analysis. The control variables are: logarithm of firm age, logarithm of firm employment, number of subsidiaries,
logarithm of number of shareholders, peer group employment, logarithm of
capital per employee, logarithm of labour productivity, industry dummies
(NACE Rev 2, at 2-digit level) and country dummies.
Table 2 presents unconditional sales turnover and employment volatilities
for FOEs and DOEs for each country separately. In addition, it gives a picture of the differences between conditional and unconditional volatilities for
these two groups of enterprises. It can be observed that for the majority of
countries unconditional sales turnover and employment volatilities are higher
in FOEs than in DOEs. However, this is not a uniform result, since these differences are negative and statistically significant for several countries: turnover volatility is statistically significantly higher among domestic firms in
France, Greece, Spain, the Czech Republic and Hungary, while employment
volatility is higher among domestic firms in Greece and Spain. (Note that the
Amadeus dataset is not a random sample and the estimated unconditional
volatilities may not be representative of the whole population of firms.)
The estimation of conditional volatilities enables us to compare FOEs and
DOEs with similar characteristics. The estimated figures presented in Table 2
imply that FOEs tend to have larger employment volatility than similar
DOEs. The difference in the volatility of sales turnover in favour of FOEs is
significantly positive for 11 countries out of the 19 for which these estimates
could be assessed. (We could not apply propensity score matching for some
countries as there was an insufficient number of observations and a lack of
common support for matching.) The employment volatility is statistically
significantly higher in FOEs than in DOEs in 10 countries out of the 19.
There is only one country, Greece, where this relationship is the other way
around, i.e. the conditional volatilities of sales turnover and employment are
statistically significantly higher among DOEs than among FOEs.
ECB Working Paper 1704, August 2014
15
Table 2: Unconditional and conditional volatilities by countries: Subsidiaries
of foreign multinationals vs. domestic firms
FOE
Unconditional volatility
DOE
Difference
(FOE – DOE)
Conditional volatility
Difference after
No. of obs.
matching
(FOE – DOE)
Volatility of sales turnover
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Italy
Netherlands
Norway
Portugal
Spain
Sweden
UK
Bulgaria
Czech Rep.
Estonia
Hungary
Latvia
Lithuania
Poland
Romania
Slovakia
Slovenia
0.227
0.354
0.223
0.396
0.35
0.291
0.375
0.368
0.338
0.442
0.301
0.439
0.417
0.388
0.642
0.388
0.549
0.444
0.664
0.515
0.416
0.891
0.501
0.379
0.217
0.319
0.233
0.39
0.368
0.251
0.432
0.37
0.299
0.433
0.337
0.453
0.384
0.374
0.611
0.411
0.564
0.472
0.671
0.502
0.362
0.67
0.444
0.381
0.010
0.035+
−0.010
0.005
−0.019+
0.040+
−0.057+
−0.002
0.039+
0.010
−0.036
−0.014+
0.033+
0.014+
0.031+
−0.024+
−0.016
−0.028+
−0.007
0.013
0.054+
0.221+
−0.003
0.008
200
7115
4002
4075
6006
4463
1459
16730
2520
23331
1014
91612
16138
24459
1502
3525
2060
148
1262
2231
11117
679
58
2087
0.042*
0.029*
0.016*
0.011
−0.009
0.036*
−0.056*
0.034*
−0.011
0.019*
−0.017
0.010
0.029*
0.017*
682
7116
4211
3853
5453
3867
1464
15990
2273
17611
656
90395
16169
24323
0.046*
0.005
0.033*
−0.011
0.045*
−0.047*
0.018*
0.046*
0.048*
−0.007
0.041*
0.045*
0.013*
0.022
−0.015
0.002
−0.001
0.019
0.041*
0.161*
Volatility of employment
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Italy
Netherlands
Norway
Portugal
Spain
Sweden
UK
ECB Working Paper 1704, August 2014
0.187
0.25
0.162
0.265
0.239
0.194
0.067
0.36
0.285
0.295
0.18
0.286
0.324
0.281
0.182
0.225
0.153
0.264
0.248
0.159
0.120
0.323
0.27
0.285
0.197
0.298
0.308
0.26
0.005
0.024+
0.010
0.0004
−0.009
0.035+
−0.053+
0.037+
0.015
0.009
−0.017
−0.012+
0.016+
0.020+
16
Conditional volatility
Difference after
No. of obs.
matching
(FOE – DOE)
Bulgaria
0.461
0.445
0.016
−0.017
1523
Czech Rep.
0.318
0.287
0.031+
0.038*
3378
Estonia
0.311
0.317
−0.006
−0.006
2003
Hungary
0.157
0.208
−0.051
79
Latvia
0.332
0.338
-0.005
−0.01
1241
Lithuania
0.35
0.317
0.033+
0.012
2233
Poland
0.245
0.189
0.056+
0.033*
10778
Romania
0.446
0.399
0.047+
0.039
680
Slovakia
0.353
0.359
−0.006
58
Slovenia
0.242
0.251
-0.01
−0.005
2180
Notes: Volatility is estimated as a coefficient of variation (CV) over the years 2001–2009,
control variables are from 2005. Firms with fewer than 5 observations are excluded, except
for Denmark where firms with a minimum of 4 observations were used. Conditional volatilities are not estimated for some countries due to the small sample size. + indicates statistical
significance of the difference in unconditional volatility (based on a t-test) at the 5% level of
significance. * indicates statistical significance of the difference in conditional volatility at
the 5% level of significance based on bootstrapped standard errors.
FOE
Unconditional volatility
DOE
Difference
(FOE – DOE)
Next, we compare sales turnover and employment volatilities for two subsets of the pooled datafile: Western European and Central and Eastern European countries.7 These two groups are differentiated throughout the paper as
the income levels and institutional backgrounds differ substantially between
these country groups. We discuss the institutional differences in more detail
in Section 5. In addition, we assess volatility separately for services and
manufacturing companies. The estimated volatilities presented in Table 3 are
indicative of the existence of the following regularities or “stylised facts”.
First, volatility of sales turnover is larger than volatility of employment. (This
is a standard result in the related literature which can be explained by inelastic labour demand.) Second, unconditional volatilities of sales turnover and
employment are higher in services than in manufacturing. Third, conditional
on firm characteristics, both sales turnover and employment are more volatile
in the subsidiaries of foreign multinationals than in domestically owned companies.8
7
WE countries are: Belgium, Finland, France, Germany, Italy, the Netherlands, Norway,
Portugal, Spain, Sweden and the UK. CEE countries are: Bulgaria, the Czech Republic, Estonia, Poland, Romania, Slovakia and Slovenia. The same groups of countries are used in the
forthcoming section on labour demand equations.
8
Although it is not the aim of this paper to compare multinationals with domestic and
foreign owners, we can still distinguish these groups in our data. The conditional employment volatility is higher among foreign-owned multinationals than among domestically
ECB Working Paper 1704, August 2014
17
Table 3: Unconditional and conditional volatilities by country groups:
Subsidiaries of foreign multinationals vs. domestic firms
Unconditional volatility
FOE
DOE
Difference
(FOE–DOE)
Conditional volatility
Difference
after
matching
(FOE–DOE)
No.
Of
obs.
Volatility of sales turnover
E Manufacturing
WE Services
WE difference
(services – manufacturing)
CEE Manufacturing
CEE Services
CEE difference
(services – manufacturing)
WE Manufacturing
0.336
0.428
0.344
0.449
−0.008+
−0.022+
0.024*
0.037*
47124
152066
0.092+
0.441
0.503
0.105+
0.384
0.449
0.057+
0.054+
0.031*
0.037*
7486
14048
0.062+
0.065+
Volatility of employment
0.236
−0.0003
0.023*
45705
−0.004+
0.021*
143462
0.062+
0.081+
0.034*
0.030*
7362
13745
0.236
WE Services
0.302
0.306
WE difference
(services – manu0.066+
0.070+
facturing)
CEE Manufacturing
0.285
0.224
CEE Services
0.326
0.245
CEE difference
(services – manu0.041+
0.021+
facturing)
Notes: See notes for Table 2 and footnote no 6.
The results for unconditional and conditional volatility are somewhat different in the groups of WE and CEE countries. The FOEs are less volatile
than DOEs in WE countries before firm characteristics are controlled for and
this difference reverses to become positive after the control for firm characteristics. On the other hand, foreign-owned firms are more volatile than domestically owned firms before and after firm characteristics in CEE are controlled for and the difference in volatility diminishes by roughly half after
matching. A possible reason for these diverging outcomes is that foreign
firms have somewhat different characteristics in WE and CEE, and also that
owned multinationals in manufacturing, while the conditional difference is not statistically
significant or becomes negative in services.
ECB Working Paper 1704, August 2014
18
foreign firms operate in less volatile industries in WE and in more volatile
areas in CEE. This finding is in accordance with the implications from the
theoretical literature (Cunat and Melitz (2012)) that more flexible labour
market institutions in CEE may attract more volatile FDI.
Appendix 2 presents the probit models behind these propensity score estimates. The appendix shows that the “propensity to be a foreign-owned firm”
is often different in WE and CEE in terms of industry variables, meaning
there are differences in the concentration of FDI to certain industries. For
example there is relatively more FDI in labour-intensive manufacturing industries in the CEE countries (textiles and wearing apparel, wood products
and furniture manufacturing) and in some volatile manufacturing industries
(non-metallic mineral products, fabricated metal products, electrical equipment, and motor vehicle manufacturing). The electrical equipment industry is
one of the largest in the sample and one of the most volatile, like it is in the
study of Cunat and Melitz (2012).
Second, these country groups differ in the conditional employment volatility of foreign firms. While there are hardly any differences in conditional
turnover volatility between WE and CEE, the difference in conditional employment volatility is somewhat higher among foreign firms in CEE than
foreign firms in WE. A “similar” foreign firm has 7–8% higher sales turnover
volatility in WE than a DOE does and 8% higher sales turnover volatility in
CEE, whereas a “similar” foreign firm has 7–10% higher employment volatility in WE and 12–15% higher employment volatility in CEE. This indicates
that foreign firms are more prone to volatile employment in CEE than in WE.
The following section will investigate whether differences in labour demand elasticity could explain the higher employment volatility of foreign
firms.
5. Elasticity of labor demand
5.1. Estimation methodology
We estimate the following labour demand equation, assuming that capital
is fixed in the short-run and that employment is adjusted on a given output, yit
(a similar approach to Barba Navaretti (2003); and Görg et al. (2009)):
lit   0  1lit 1  1wit   2 yit   t   s   it
(11)
where lit is log(employment) in firm i at time t (t = 1, …,9); wit is log(real
labour cost per employee); yit is log(real output); τt notes time dummies and
ECB Working Paper 1704, August 2014
19
γs sector dummies (NACE 2-digit industries). Estimations covering the data
from multiple countries include time dummies for each country, i.e.
time*country dummies. Sector dummies are included in the base specification. However, for some estimations sector dummies were excluded when
specification tests indicated poor fit of the specification or unfeasible coefficients were produced. Nominal variables are deflated by 2-digit industry level GDP deflators to obtain real values, see also the discussion in the data section. The coefficient α1 captures firms’ employment persistence (speed of
adjustment = 1 – α1). The coefficient β1 measures short-term wage elasticity
of labour demand and β2 short-term output elasticity of labour demand. Longterm elasticities can be found by dividing short-term elasticities by the speed
of adjustment.
We introduce the interaction terms with foreign ownership to test for the
differences in the labour demand elasticities of domestic and foreign firms:
lit   0  1lit 1  1wit   2 yit
  2 FOi  lit 1  3 FOi  wit   4 FOi  yit   t   s   it
(12)
where FOi takes the value “1” when a company is foreign-owned and the
value “0” when a company is domestically owned. Coefficients of the interactive variables capture the differences between FOEs and DOEs in employment persistence and short-term labour demand elasticities. If the speed of
employment adjustment is higher in FOEs than in DOEs, we will observe the
coefficient α2 to be negative and statistically significant. If the short-term
wage elasticity of labour demand is higher in absolute terms for FOEs, we
will observe coefficient β3 to be negative and statistically significant. Similarly, if the short-term output elasticity of labour demand is higher in FOEs than
in DOEs, β4 will be positive and statistically significant.
5.2. Elasticity of labour demand: Differences between FOEs
and DOEs across countries
Regression equation (12) is estimated by the system GMM method9 developed by Arellano and Bover (1995) and Blundell and Bond (1998). We
9
OLS and fixed effects (FE) estimations were also carried out. FE estimates are biased in
dynamic panels (Nickell (1981)). Since employment and its lagged value are positively correlated, the FE estimate for the lagged dependent variable is downward biased. This also
implies that the OLS estimate of the coefficient on the lagged dependent variable is upward
biased. Thus the OLS and FE estimates of the lagged term determine a lower and upper
bound for the estimated speed of adjustment. Note that the same boundaries could be applied
for the other control variables included in the model only under assumption of their exogene-
ECB Working Paper 1704, August 2014
20
employ a two-step system GMM estimation with Windmeijer-corrected
standard errors.10 The lagged employment and real turnover are treated as
endogenous variables in the model; real wages are treated as endogenous,
pre-determined or exogeneous dependent on the coefficients and specification tests. We choose the dynamic form of our labour demand equation and
the set of instruments from the serial correlation tests (Arellano and Bond
(1991)) and the Hansen test for overidentifying restrictions (Hansen (1982)).
We imply Hansen’s test for overidentifying restrictions for testing the validity of the joint set of instruments. As is usual for system GMM estimations,
the overidentification tests tend to reject the null hypothesis of no overidentification in large and heterogeneous samples. Arellano and Bond (1991) show
that rejection takes place too often in the presence of heteroskedasticity. Our
pooled sample of all countries is relatively large, which increases the probability that the tests of overidentifying restrictions are subject to type I error.
The tests for second-order serial correlation are also subject to the criticism
that they are inclined to type I error in samples with large cross-sections relative to the time dimension.
OLS and fixed effects (FE) estimations were also carried out to assess the
sensitivity of the estimated coefficients to the various estimation techniques.
The estimated coefficients for other explanatory variables (except for the
lagged dependent variable) tend to be between the OLS and FE for wages
and output, and are often larger than the OLS and FE for ownershipinteracted wages and output. The endogeneity of wage and output against
employment in DOEs and FOEs should be accounted for by the system
GMM estimation as most of the Hansen tests applied to our regressions do
not reject the null hypothesis of no overidentification of instruments.
Our first choice for the dynamic form is that specified in equation (12). If
the specification tests described above reject the assumption of no secondorder autocorrelation or the validity of instruments, or the coefficient of the
lagged dependent variable does not lie within the brackets of fixed effects
and OLS estimation, we use the specification where the second lag of the
dependent variable is added to the RHS. Since the time dimension of the
sample is 9 years at maximum, we include at most 2 lags of the dependent
variable. If the specification tests and OLS and FE brackets are not satisfied
for this dynamic form either, the third specification adds the first lag of wages and output to the RHS. As a result the applied dynamic form varies from
country to country.
ity, which in our specification is not valid. See Bond (2002) for this discussion. Difference
GMM is not used in this paper as employment, output and wages are highly persistent timeseries and hence their levels provide weak instruments for differences.
10
We use the xtabond2 command for Stata, see Roodman (2009).
ECB Working Paper 1704, August 2014
21
We also experimented with various sets of instruments and could not find
a common set of instruments that would have been suitable for all countries.
The differences in dynamic form and the set of instruments arise from different properties of the time-series across countries, cross-country differences in
the time-dimension and object-dimension of the panel, and possibly also
from differences in the institutions that shape the endogeneity of the explanatory variables.
We start out by estimating the labour demand relationship as specified in
equation (12) separately for each country. Only firms with at least 5 consecutive observations for employment, wages and output, and without any gap in
these series are included in the estimation sample. Firms that show yearly
growth of 100% or more in employment, wages or output are excluded and
taken as measurement error or merger/acquisition, which we cannot control
for. There are 18 countries covered in this and the following sub-sections.
The estimated effects for the interactive variables imply whether the elasticity
of labour demand is different for FOEs and DOEs in each country. The estimated coefficients for specification (12) are presented in Tables 1 and 2 in
Appendix C. Estimates for the interactive variables capturing the differences
between short-term wage and output elasticities and speed of adjustment are
insignificant for the majority of the countries covered. However, when the
estimates indicate a faster speed of adjustment for foreign firms, it is always
accompanied by greater (absolute) wage and output elasticity, while slower
speeds come with lower elasticity. Consequently, all three indicators imply
either greater or lower flexibility of labour adjustment for foreign firms.
Appendix C indicates that the speed of adjustment of foreign firms is statistically significantly higher in manufacturing in Italy and Slovenia and in
services in Portugal and Bulgaria. The opposite is found in manufacturing in
France and services in the Netherlands. The estimated coefficients on
FO*log(rwage) are statistically significantly negative (implying larger elasticity in absolute terms in FOE) for manufacturing in Belgium and Italy,
whereas they are statistically significantly positive for services in Finland and
the Netherlands. The short-term output elasticity of labour demand is statistically significantly lower for foreign firms in manufacturing in France and in
services in Finland and the Netherlands. Thus, country-by-country regressions do not yield conclusive results for the difference in labour demand between domestic and foreign companies. Grouping countries together in the
groups of Western Europe and Central and Eastern Europe as in the previous
section does not reveal any differences in foreign or domestic firms either
(see Table 4).
ECB Working Paper 1704, August 2014
22
Table 4: Labour demand estimates of FOEs and DOEs, 2001–2009: country
groups
L.log(empl)
L2.log(empl)
Log(rwage)
L.Log(rwage)
Log(rturn)
L.Log(rturn)
L.FO* log(empl)
L2.FO* log(empl)
FO*log(rwage)
L. FO*log(rwage)
FO*log(rturn)
L. FO*log(rturn)
Sector dummies
Year*country dummies
# of obs.
# of groups
Min obs. gr.
Mean obs. gr.
Max obs. gr.
# of instruments
Hansen p
AR(1) test
AR(2) test
FDI in sample
Western Europe
Central and Eastern Europe
Manufacturing
Services
Manufacturing
Services
GMM SYS (3 .) GMM SYS (3 5) GMM-SYS (3 .) GMM SYS (3 .)
wage pre
wage pre
wage pre
wage ex
0.853***
0.611***
0.856***
0.737***
(0.081)
(0.153)
(0.101)
(0.218)
0.011
(0.077)
−0.546***
−0.382***
−0.291**
−0.675***
(0.085)
(0.136)
(0.125)
(0.242)
0.461***
(0.082)
0.654***
0.250**
0.274***
0.504***
(0.064)
(0.103)
(0.087)
(0.168)
−0.534***
(0.085)
−0.073
0.207
−0.014
−0.082
(0.112)
(0.134)
(0.092)
(0.265)
0.087
(0.100)
0.018
0.120
0.106
0.255
(0.072)
(0.165)
(0.102)
(0.223)
−0.034
(0.073)
−0.001
−0.177
-0.067
−0.105
(0.066)
(0.134)
(0.078)
(0.188)
0.003
(0.055)
yes
yes
yes
yes
yes
yes
yes
yes
232058
718913
30648
58701
41004
114224
4945
9721
2
3
3
3
5.659
6.294
6.198
6.039
7
8
8
8
211
188
182
158
0.123
0.441
0.601
0.555
−8.056
−3.144
−7.809
−2.958
0.392
1.281
−1.192
−0.926
0.187
0.158
0.416
0.334
Notes: System GMM estimations. Dependent variable: log(employment), 2001–2009. Twostep estimators with Windmeijer-corrected cluster robust standard errors in parentheses.
Lagged employment and turnover are treated as endogenous; wages are treated as endogenous, pre-determined or exogenous dependent on specification tests. The range of lag lengths
of GMM type instruments is reported at the top of each column in parentheses. “FO in the
sample” refers to the share of foreign-owned companies. *, **, *** indicate statistical significance at the 10%, 5% and 1% level of significance, respectively. See footnote no 6 for the
list of host countries covered.
ECB Working Paper 1704, August 2014
23
There are even fewer statistically significant differences between domestic
and foreign firms in long-run elasticities (see Appendix D). Long-run wage
or turnover elasticity is found to be lower for foreign firms (in absolute value) in services in Finland and Spain, and higher in services in Italy. The
speed of adjustment is on average higher in services and long-run elasticities
are higher in manufacturing, which is to be expected given the smaller firm
size in services and the higher substitutability of labour in manufacturing.
The results by country groups presented in Table 4 do not indicate any significant differences between foreign and domestic firms in long-run elasticities
either.
Overall we do not find similar conclusive results for foreign firms’ higher
speed of adjustment to those found by Barba Navaretti et al. (2003). However, they used difference GMM for estimating the labour demand equations,
which might be poorly identified due to weak instruments in estimations with
highly persistent variables (see the discussion by Bond (2002)). Our results
are in line with the findings of Buch and Lipponer (2010) and Hakkala et al.
(2010), who find no statistically significant differences between the labour
demand of foreign and domestic firms in Germany and Sweden. However,
the results seem to be country-specific, as in some countries the differences
between foreign and domestic firms are large and statistically significant.
French and Spanish foreign firms, for example, seem to behave much more
inelastically than their domestic counterparts, and it is worth noting that these
countries have relatively strict employment protection legislation. The remaining sections of the paper investigate whether the differences between
domestic and foreign firms can be explained by the home and host country
labour market institutions.
5.3. Elasticity of labour demand: Labour market
institutions
This section analyses whether labour market institutions could have an effect on labour demand elasticities and whether institutions could explain the
differences in elasticities of FOEs and DOEs. We separate the sample into
domestically and foreign-owned firms and analyse how labour market institutions affect the elasticity of labour demand in the two groups. For this purpose, we introduce interaction terms with measures of labour market regulations to the labour demand equation and estimate the following specification
on two subsamples, DOE and FOE:
lit   0  1lit 1  1wit   2 yit
  2 INSTct  lit 1  3 INSTct  wit   4 INSTct  yit   t c   s   it
ECB Working Paper 1704, August 2014
(13)
24
where INSTct denotes the measure of labour market regulations in country c
at time t and ηc denotes a set of country dummies.
We include two measures of labour market regulations in the regressions:
union density, which is based on statistics from the OECD and ICTWSS database by Visser (2011), and the OECD’s employment protection legislation
(EPL) index (Version 2 published in 2009).11 Appendix E presents the average values of these measures for 2001–2009 across the countries covered and
the USA. Despite significant differences in income and wage levels within
Europe (see Appendix F), the strictness of employment protection legislation
does not diverge much across European countries according to the OECD
measure. The UK stands out with a low value for the EPL index, while Portugal and Spain have the highest EPL indices in Europe. The EPL index reflects formal regulations. However, there is evidence that the actual labour
market flexibility is higher in CEE due to weak enforcement of EPL (Eamets
and Masso (2005)). To show a picture of the institutional differences in the
home and host countries of MNEs, we present the weighted average
measures of EPL and union density for the home countries of foreign subsidiaries operating in each country in Appendix D.
We interpret both EPL and union density as proxies of labour market
strictness. High union coverage is associated with more staggered employment adjustments and should lead to less elastic labour demand. We include
interactive country-year dummies in the regressions as additional controls for
country-specific time trends capturing any other country-specific developments that may affect the elasticity of labour demand.
We present the results separately for countries from Western Europe and
those from Central and Eastern Europe as the enforcement of institutions
could differ between these country groups and the overall cost of employment adjustment is different due to the vast differences in wage costs (see
Appendix E). EPL tends to be much more persistent over time than union
coverage does during 2001–2009 in Europe. Union density measures exhibit
more dynamism. The results (presented in Table 5 and 6) imply that more
strictly regulated labour markets are associated with a lower speed of adjustment, lower wage elasticity for employment and lower output elasticity for
employment among domestic firms, as could be expected. Union density declined in most countries and employment contracts become less strictly regulated in 2001-2009, although changes in EPL were less pronounced. Given
these trends, the estimated coefficients imply that the reduction in the strict-
11
Our preferred measure of regulations related to collective bargaining would be union
coverage. However, this measure is often missing and only irregularly available for many of
the countries that our dataset covers and therefore we use union density.
ECB Working Paper 1704, August 2014
25
ness of labour market regulations was associated with increasing elasticity of
labour demand in 2001–2009.
Both of the measures we use (union density and EPL) yield similar results
for domestic firms, since these two forms of labour market regulation tend to
be complements: European countries that generally have more powerful unions also tend to have stricter EPL. (Please refer to the theoretical model developed by Bertola and Rogerson (1997) for an explanation of why these two
institutions should be complements.) It is worth noting that EPL has a statistically significant effect on domestic firms’ labour demand in WE, while union density has a statistically significant effect on labour demand in the CEE
countries. In Western European countries, our measure of union power (union density) may yield insignificant results because it is not sufficiently correlated with the actual coverage of collective bargaining. This is less of a
problem in the CEE countries since union agreements are not typically extended to non-union members, as is customary in several WE countries (such
as France, Italy, and Spain), and therefore collective bargaining coverage and
trade union membership have an almost one-to-one correspondence in CEE.
On the other hand, the OECD’s EPL index may be a better measure of the
actual strictness of labour regulations in WE than in CEE due to better enforcement of labour regulations in WE. In conclusion, the insignificance of
the estimated effects may stem from measurement errors in the indicators of
the labour market institutions that we employ. When variables are measured
with errors then the estimated effects tend to be biased towards zero.
The estimated results imply that a stricter regulatory environment is associated with less elastic labour demand for domestic firms. Surprisingly, the
foreign firms’ reaction to host country institutions is different in WE and in
CEE. While foreign firms in WE tend to behave even more elastically in the
presence of stricter labour market institutions, foreign firms in CEE have less
elastic labour demand in a stricter institutional environment. There is no good
theoretical explanation for the estimated effects for WE. One possible explanation is that FDI in WE and CEE have different motivations and characters.
Another explanation is that as the sample of foreign-owned companies in WE
is dominated by companies hosted by the UK and originating from the US
(see also Figure 1), the more inelastic US firms in the UK, with its relatively
weak EPL, are distorting the relationship. If the UK is removed from the
sample of foreign firms in WE, the statistically significantly negative effect
of the host institutions disappears. This specific case illustrates the importance of also controlling for home country institutions in the estimations
of host country effects, as we do in the following estimations (Tables 5 and
6).
ECB Working Paper 1704, August 2014
26
Table 5: Labour market institutions and the elasticity of labour demand,
manufacturing 2001–2009, dependent variable: log(employment)
FOEs in WE
DOEs in WE
L.lempl
EPL (3 .)
wage pre
0.759***
(0.117)
UD (3 .)
wage pre
0.855***
(0.092)
EPL (3 4)
wage pre
0.792***
(0.109)
UD (3 5)
wage pre
0.740***
(0.080)
−0.229**
(0.103)
−0.137**
(0.069)
−0.174
(0.131)
−0.240***
(0.077)
−0.211***
(0.062)
−0.104**
(0.045)
0.211**
(0.095)
0.176**
(0.069)
0.163*
(0.093)
0.284***
(0.056)
0.215***
(0.049)
0.125***
(0.029)
−0.004
(0.053)
−0.382***
(0.141)
−0.043
(0.036)
0.008
(0.064)
0.017
(0.022)
−0.002
(0.006)
0.001
(0.069)
0.128
(0.087)
0.030
(0.060)
−0.109
(0.080)
yes
UD (3 .)
wage pre
0.947***
(0.109)
−0.018
(0.103)
−0.507***
(0.086)
0.410***
(0.092)
0.590***
(0.090)
−0.512***
(0.121)
−0.284
(0.183)
0.056
(0.198)
−0.183
(0.149)
0.180
(0.188)
0.125
(0.147)
0.060
(0.209)
yes
Ratio of host and home
institutions
EPL (3 5) UD (2 .)
wage pre
0.828***
0.842***
(0.072)
(0.050)
−0.004
(0.054)
−0.296**
(0.117)
−0.045
(0.054)
0.170*
(0.097)
0.057*
(0.031)
−0.017
(0.014)
0.016
(0.048)
0.201*
(0.121)
0.039
(0.042)
−0.121*
(0.072)
−0.041*
(0.023)
0.011
(0.010)
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
222483
33395
3
6.662
8
167
0.597
−13.578
2.523
2.664
188720
33395
2
5.651
7
209
0.897
−5.774
0.579
0.259
51021
7609
3
6.705
8
212
0.063
−9.316
−1.915
2.165
51021
7609
3
6.705
8
212
0.031
−8.020
−1.877
0.282
49234
7346
3
6.702
8
168
0.225
−5.520
−1.457
1.795
49460
7377
3
6.705
8
188
0.285
−7.910
−1.858
0.253
49234
7346
3
6.702
8
204
0.442
−9.707
−1.876
1.595
49460
7377
3
6.705
8
256
0.214
−11.007
−1.712
1.602
EPL (3 4)
wage pre
0.621***
(0.154)
L2.lempl
lrwage
L.lrwage
lrturn
L.lrturn
L.INST*lempl
−0.637***
(0.184)
0.334
(0.232)
0.626***
(0.175)
−0.359*
(0.213)
0.121**
(0.056)
L2.INST*lempl
INST*lrwage
L.INST*lrwage
INST*lrturn
L.INST*lrturn
Sector dummies
Year*country
dummies
# of obs.
# of gro~s
Min.. gr.
Mea.. gr.
Max.. gr.
# of instr
Hansen p
AR(1)
AR(2)
INST in sample
ECB Working Paper 1704, August 2014
Host institutions
Home institutions
27
Table 5 (continued).
FOEs in CEE
DOEs in CEE
EPL (2 .)
L.lempl
L2.lempl
lrwage
lrturn
L.INST*empl
L2.INST*empl
INST*lrwage
INST*lrturn
Sector dummies
Year*country
dummies
# of obs.
# of gro~s
Min.. gr.
Mea.. gr.
Max.. gr.
# of instr
Hansen p
AR(1)
AR(2)
INST in sample
Host institutions
UD (2 .)
EPL (3 .)
wage pre
0.889*** 0.813***
(0.067)
(0.130)
0.997***
(0.080)
−0.084**
(0.035)
−0.156
−0.110
(0.112)
(0.082)
0.173
0.139*
(0.115)
(0.074)
−0.008
0.139
(0.012)
(0.098)
0.005
(0.011)
0.027
0.220**
(0.043)
(0.110)
−0.024 −0.165**
(0.041)
(0.083)
yes
yes
UD (2 .)
Ratio of host and
home institutions
UD (3 .) EPL (3 5)
UD (3 5)
wage pre wage pre
wage pre
0.755*** 0.733***
0.731***
(0.097)
(0.135)
(0.112)
Home institutions
0.839***
(0.074)
EPL (3 5)
wage pre
0.760***
(0.102)
−0.341*
(0.178)
0.393*
(0.233)
0.017
(0.0z37)
−0.226***
(0.065)
0.223***
(0.058)
0.169
(0.140)
−0.111
(0.158)
0.216**
(0.089)
0.020
(0.024)
−0.257**
(0.104)
0.267***
(0.062)
0.008
(0.180)
−0.331***
(0.126)
0.302***
(0.086)
0.012
(0.009)
−0.306***
(0.112)
0.288***
(0.081)
0.020
(0.015)
0.063
(0.072)
−0.085
(0.098)
yes
0.266**
(0.135)
−0.220*
(0.120)
yes
−0.045
(0.056)
0.017
(0.033)
yes
0.060
(0.208)
−0.035
(0.143)
yes
0.029
(0.040)
−0.017
(0.021)
yes
0.043
(0.027)
−0.028*
(0.016)
yes
yes
yes
yes
yes
yes
yes
yes
yes
14922
2953
2
5.053
7
251
0.027
−8.911
−0.119
2.277
17890
2953
3
6.058
8
258
0.975
−10.009
−1.502
0.241
12758
1992
3
6.405
8
182
0.211
−7.762
−0.019
2.174
12758
1992
3
6.405
8
258
0.327
−10.110
0.577
0.219
11057
1725
3
6.410
8
182
0.186
−6.841
−0.231
2.177
11173
1741
3
6.418
8
182
0.799
−6.585
−0.232
0.295
11057
1725
3
6.410
8
158
0.466
−4.421
−0.864
1.174
11173
1741
3
6.418
8
158
0.680
−5.615
−0.662
1.083
Notes: See notes for Table 4 and footnote no 6 for the list of host countries covered. EPL
denotes OECD employment protection legislation index and UD union density. „INST in
sample“ refers to an average value of the institutional measure (EPL or UD).
ECB Working Paper 1704, August 2014
28
Table 6: Labour market institutions and the elasticity of labour demand, services 2001–2009, dependent variable: log(employment)
FOEs in WE
DOEs in WE
L.lempl
lrwage
L.lrwage
lrturn
L.lrturn
L.INST*lempl
INST*lrwage
L.INST*lrwage
INST*lrturn
L.INST*lrturn
Sector dummies
Year*country
dummies
# of obs.
# of gro~s
Min.. gr.
Mea.. gr.
Max.. gr.
# of instr
Hansen p
AR(1)
AR(2)
INST in sample
EPL (2 3)
wage pre
0.854***
(0.113)
−0.485***
(0.114)
0.389***
(0.125)
0.766***
(0.148)
−0.576***
(0.123)
0.020
(0.029)
−0.008
(0.032)
0.093**
(0.043)
−0.120**
(0.052)
0.081**
(0.036)
yes
Host institutions
Home institutions
UD (3 .)
wage pre
0.711***
(0.109)
EPL (3 4)
wage ex
0.893***
(0.253)
−0.382
(0.502)
UD (2 4)
wage pre
0.666***
(0.146)
−0.083
(0.107)
EPL (2 4)
wage pre
0.680***
(0.103)
−0.341*
(0.197)
−0.124
(0.089)
0.111
(0.409)
0.113
(0.102)
0.194*
(0.104)
0.249***
(0.084)
−0.110
(0.067)
−0.017
(0.077)
0.029
(0.181)
0.079
(0.134)
−0.071
(0.145)
0.007
(0.031)
0.021
(0.055)
0.034
(0.102)
−0.082
(0.192)
0.034
(0.117)
0.003
(0.042)
yes
yes
yes
yes
0.756***
(0.092)
−0.567***
(0.095)
0.404***
(0.081)
0.500***
(0.085)
−0.443***
(0.098)
0.235*
(0.124)
0.017
(0.130)
0.124
(0.139)
−0.011
(0.089)
−0.125
(0.113)
yes
yes
yes
yes
yes
yes
605306
96540
3
6.270
8
186
0.227
−13.681
−0.985
2.644
605306
96540
3
6.270
8
212
0.060
−7.046
−0.228
0.276
113607
17684
3
6.424
8
146
0.977
−2.400
−1.032
2.145
113607
17684
3
6.424
8
211
0.136
−4.823
−1.902
0.312
106897
16655
3
6.418
8
211
0.500
−6.175
−1.010
1.841
ECB Working Paper 1704, August 2014
UD (2 4)
Ratio of host and
home institutions
UD (3 5)
EPL (3 .)
wage pre
0.744*** 0.743***
(0.078)
(0.108)
−0.131
−0.270**
(0.116)
(0.135)
0.043
(0.067)
0.164*
(0.087)
−0.010
(0.022)
−0.016
(0.036)
−0.014
(0.011)
0.007
(0.018)
0.031
(0.024)
0.003
(0.015)
yes
yes
yes
yes
yes
107500
16740
3
6.422
8
227
0.382
−8.605
−1.977
0.252
106897
16655
3
6.418
8
239
0.045
−8.241
−2.071
1.545
107500
16740
3
6.422
8
187
0.042
−5.195
−1.485
1.818
29
Table 6 (continued).
FOEs in CEE
DOEs in CEE
EPL (3 4)
L.lempl
lrwage
lrturn
L.INST*lempl
INST*lrwage
INST*lrturn
Sector dummies
Year*country
dummies
# of obs.
# of gro~s
Min.. gr.
Mea.. gr.
Max.. gr.
# of instr
Hansen p
AR(1)
AR(2)
INST in sample
0.792***
(0.132)
−0.296
(0.467)
0.333
(0.240)
0.009
(0.007)
−0.016
(0.180)
−0.053
(0.103)
yes
Host institutions
UD (3 4) EPL (2 .)
wage ex
0.653*** 0.764***
(0.230)
(0.099)
−0.395
−0.108
(0.592)
(0.078)
0.320
0.182***
(0.612)
(0.066)
0.012
−0.001
(0.023)
(0.009)
−0.089
−0.015
(0.295)
(0.014)
0.031
0.002
(0.228)
(0.017)
yes
yes
Home institutions
UD (2 .)
EPL (2 .) UD (2 4)
wage pre
wage pre
0.597*** 0.731*** 0.646***
(0.129)
(0.132)
(0.126)
−0.204** −0.268
−0.299*
(0.092)
(0.167)
(0.172)
0.234*** 0.219**
0.299***
(0.072)
(0.107)
(0.099)
0.212*
−0.016
0.308*
(0.121)
(0.021)
(0.182)
0.066
0.004
−0.012
(0.117)
(0.054)
(0.174)
−0.019
0.001
−0.067
(0.124)
(0.035)
(0.116)
yes
yes
yes
Ratio of host and
home institutions
EPL (2 .) UD (2 3)
wage pre
0.799*** 0.733***
(0.099)
(0.190)
−0.138
−0.230
(0.087)
(0.248)
0.161***
0.205
(0.059)
(0.126)
−0.018
0.009
(0.011)
(0.019)
−0.011
−0.020
(0.022)
(0.029)
0.008
0.009
(0.012)
(0.018)
yes
yes
yes
yes
yes
yes
yes
yes
yes
yes
39116
6518
3
6.001
8
150
0.068
−5.047
−0.540
2.171
39113
6517
3
6.002
8
117
0.369
−1.908
−1.314
0.216
19588
3204
3
6.114
8
258
0.406
−6.583
0.251
2.174
19588
3204
3
6.114
8
226
0.787
−4.629
0.085
0.213
17285
2810
3
6.151
8
258
0.873
−4.554
−0.308
2.097
17374
2823
3
6.154
8
182
0.976
−5.743
−1.018
0.292
17285
2810
3
6.151
8
258
0.842
−6.350
−0.201
1.268
17374
2823
3
6.154
8
156
0.874
−3.422
−1.039
1.099
Notes: See notes for Table 4 and footnote no 6 for the list of host countries covered. EPL
denotes OECD employment protection legislation index and UD union density.
Differences in the elasticity of labour demand between FOEs and DOEs
could be influenced by institutional differences in the home and host countries of multinationals. Table 5 and Table 6 test for the relevance of home
country institutions in MNEs’ labour demand. These results are more consistent across country groups and imply that FDI from countries with stricter
labour market regulations tends to have less elastic labour demand. This result could be interpreted as an indication of spillover effects of institutions
from home to host countries within firms. However, this interpretation may
not be valid, as the decision to invest in a particular country is subject to both
home and host institutions, and we are not controlling for host country institutions in these regressions.
To address these concerns we introduce a variable which is the ratio of the
measures of labour market regulations (EPL index, union density) in the host
and home countries. This ratio is calculated for each subsidiary of a foreignowned company and is variable over time and across all bilateral pairs of
ECB Working Paper 1704, August 2014
30
home-host relationships. The decision to invest in a company in a particular
country might be motivated by the difference in host and home institutions.
Firms in countries with strict regulations might look for investments in countries with weak regulations to reduce the costs of employment adjustment
caused by demand volatility. Our results confirm this hypothesis; the institutional difference is statistically significant in manufacturing and the interaction terms indicate that the stricter the home country institutions are relative
to those of the host country, the more elastic the labour demand is in the foreign-owned subsidiary of an MNE in the host country compared to the demand of other MNEs. This regularity also holds in the opposite direction: the
weaker the home country institutions are relative to those of the host country,
the less elastic the labour demand of MNEs is as it is less costly for them to
adjust for employment changes in their home country.
The relative distances between the measures of host and home country institutions can explain only a small portion of the difference in labour demand
elasticities between FOEs and DOEs. This result is at least partly caused by
the use of measures which do not capture well the actual differences in institutions. The OECD’s EPL index is based on formal legislation, which does
not take account of the fact that law enforcement differs between countries.
Labour market flexibility depends on norms and cultural attitudes in addition
to formalised rules, and so the EPL index, which is a combination of different
legislative procedures, is only a crude measure of the actual strictness of regulations. Union density is also a poor measure for capturing variations in actual union power across countries. Collective bargaining coverage would be a
better measure but unfortunately the complete time series are not available
for this variable for all the countries that our sample covers and so we could
not use it.
The inclusion of country-level variables for the firm-level regression estimations together with the country-time interactions means that the effect of
institutions could also be picked up by these dummies. In this context it is
relevant that we can still observe statistically significant effects for institutions in addition to the country-specific time trends. However, because of the
measurement problems discussed above, the variables that we employ have
insufficient variation and do not capture the actual differences in labour market regulations to a full extent. Therefore we also carried out additional estimations which should capture the effect of institutions. These estimations,
which are presented in the following section, can be considered as an additional consistency check to the empirical findings described above.
ECB Working Paper 1704, August 2014
31
5.4. Estimations of two subsamples
We hypothesised that institutional differences in the home and host countries matter for labour demand elasticity of multinational enterprises since
they can shift the adjustment of labour in response to economic shocks to
countries where it is easier to make the adjustment. It may be expected that
this occurs only when the institutional framework is substantially different in
the home and host countries, and so the impact of any such reallocation of
adjustment should be more prevalent when the sample is restricted to a subset
of firms for which these institutional differences are more pronounced. In
order to see whether this is the case, we evaluate the elasticities of labour
demand for two subsets of our sample. First, the subsidiaries of the US companies are compared with domestically owned firms in Western European
(WE) countries. The US labour market institutions are substantially less strict
than those of Western Europe, see Appendix E. Second, the subsidiaries of
German firms are compared with domestic companies in the Central and
Eastern European (CEE) countries. Germany’s EPL index and union density
is not significantly higher than those of the CEE countries (Appendix E), but
as noted earlier there could be more substantial differences in the enforcement of the employment regulations (Eamets and Masso (2005)). Both of
these groups represent the most important country of origin among foreign
companies as US companies make up 25% of all the foreign companies in the
WE sample and German companies make up 21% of all the foreign companies in the CEE sample.
Our first exercise focuses on subsidiaries of foreign MNEs from a country
with mostly unregulated labour markets, the USA, in a group of countries
with relatively strict labour market institutions, Western Europe.12 The results
are presented in Table 7. The estimated figures indicate that in comparison to
domestic companies, the subsidiaries of the US multinationals in Western
Europe have more persistent labour adjustment. This implies that when the
country of origin has a less regulated labour market environment, the subsidiaries of an MNE have less elastic labour demand than local companies in
their host countries as it is less costly for the MNE to adjust labour input in
the country of origin. The effects on the long-term and short-term wage and
output elasticities of labour demand are not statistically significant.
12
Franco, C. (2013) argues that as there is no substantial technological gap between the
USA and the OECD countries, the US resource-seeking FDI in OECD countries is not looking for natural resources or cheap labour but is instead looking for technological resources
that could complement or augment the resources at home.
ECB Working Paper 1704, August 2014
32
Table 7: Labour market institutions: Estimations for two subsamples, 2001–
2009, dependent variable: log(employment)
US FDI to Western Europe
L.lempl
lrwage
lrturn
L.fdiempl
fdiwage
fditurn
Sector dummies
Year*country dummies
# of obs.
# of gro~s
Min.. gr.
Mea.. gr.
Max.. gr.
# of instr
Hansen p
AR(1)
AR(2)
Share of FO in sample
Manufacturing
(lag 2 2) wage
pre
0.639***
(0.073)
−0.161*
(0.090)
0.252***
(0.063)
0.221*
(0.132)
0.028
(0.105)
−0.067
(0.089)
yes
yes
235078
35243
3
6.670
8
153
0.021
−9.252
1.102
0.054
Services
lag(3 4) wage
pre
0.580*
(0.306)
−0.487
(0.441)
0.435
(0.291)
0.125
(0.206)
0.271
(0.344)
−0.257
(0.266)
yes
yes
629588
100318
3
6.276
8
168
0.884
−2.075
−0.578
0.039
German FDI to Central and Eastern
Europe
Manufacturing
Services
(lag 3 5) wage
(lag 3 5) wage pre
pre
0.942***
0.897***
(0.110)
(0.177)
−0.178
−0.423**
(0.134)
(0.183)
0.159
0.309**
(0.108)
(0.132)
−0.406**
−0.432**
(0.186)
(0.200)
−0.195
−0.043
(0.162)
(0.199)
0.230
0.150
(0.141)
(0.151)
yes
yes
yes
yes
20759
42903
3398
7130
3
3
6.109
6.017
8
8
158
158
0.772
0.416
−7.348
−5.025
−1.737
−1.797
0.138
0.088
Notes: See notes for Table 5.
In the second case, we assess the differences in labour demand elasticities
between the subsidiaries of German firms in CEE countries and domestically
owned firms. The results are in accordance with our hypothesis in this subsample as well. The speed of adjustment is substantially higher in the subsidiaries of German-owned firms than in the local companies in CEE. This suggests that foreign subsidiaries originating from home countries with a relatively strict institutional environment have a substantially higher speed of
adjustment than domestic companies as it is more costly for the MNEs to
adjust labour inputs in their home country. The effects on the long-term and
short-term wage and output elasticities of labour demand are not statistically
significant.
ECB Working Paper 1704, August 2014
33
6. Conclusion
The purpose of the current study is to analyse how employment volatility
differs in companies with foreign and domestic owners. Our analysis is based
on an Amadeus firm-level dataset which covers more than 20 European
countries. We derive employment volatility on the basis of standard labour
supply and demand functions and demonstrate that it can be expressed as a
combination of two components. The first component captures volatility due
to changes in labour demand elasticity. Given a non-zero elasticity of labour
supply, the elasticity of labour demand is positively related with employment
volatility. The second component captures volatility in employment due to
economic shocks. The more exposed a firm is to external shocks, the higher
its employment volatility is. This decomposition indicates that the presence
of foreign-owned companies may lead to higher employment volatility because FOEs react more sensitively to wage changes in the host country or
because they are more tightly integrated in international markets and are per
se more exposed to external shocks.
The estimations of conditional volatility based on propensity score matching yield the result that employment tends to be more volatile in the subsidiaries of foreign-owned MNEs than in domestically owned firms. However,
larger volatility in foreign-owned enterprises is not unanimously caused by
their more elastic labour demand. Our estimations imply that labour demand
can be either more or less elastic in subsidiaries of foreign-owned multinationals, depending on the institutional environments of their home and host
countries. When FDI originates from a region with a more flexible institutional environment (e.g. from the USA to Western European countries) then
the elasticity of labour demand is smaller in absolute value in FOEs than in
DOEs. In the opposite case (e.g. when FDI is originating from Germany to
CEE countries) the elasticity of labour demand is higher.
A potential explanation for this finding is that in countries with rigid labour market regulations, multinational companies avoid changing domestic
employment in response to economic shocks and instead use other margins of
adjustment. They are more likely to do this than domestic firms are since it is
easier for multinational companies to substitute between factor inputs. In
addition to adjusting via alternative margins, they may also shift the adjustment of labour in response to economic shocks to subsidiaries which are located in countries with less regulated labour markets. Alternatively, multinational firms may choose the host countries where they establish subsidiaries
by looking at the labour market institutions: if they operate in sectors that
have highly volatile demand then they are more likely to move to countries
with a flexible institutional environment. In either case, the presence of foreign-owned firms would have an amplifying effect on the elasticity of labour
ECB Working Paper 1704, August 2014
34
demand in countries with flexible labour market institutions, whereas it
would have a dampening effect in countries with rigid institutions.
Due to the limitations of the Amadeus data we can only study labour adjustment via the intensive margin, i.e by assessing changes in employment in
incumbent companies. Employment may also be more volatile in foreignowned multinationals than in domestically owned firms as they are more
likely to establish and close down subsidiaries. The second of these two margins has been tested in the empirical literature and it has mostly been confirmed that FOEs are more “footloose”, i.e. they have higher conditional exit
rates, than DOEs (e.g. Bernard and Sjöholm (2003); Görg and Strobl (2003)
Alvarez and Görg (2009); Wagner and Weche Gelübke (2011)). Investigation
of the role that labour market institutions play in the entry and exit decisions
of foreign multinationals would be an interesting area for further research
that would complement the findings of the current study.
ECB Working Paper 1704, August 2014
35
References
ALVAREZ, R. AND GÖRG, H. (2009): “Multinationals and plant exit: Evidence from Chile”, International Review of Economics and Finance,
Vol. 18, pp. 45–51.
ARELLANO, M. AND BOND, S. R. (1991): “Some Tests of Specification
for Panel Data: Monte Carlo Evidence and an Application to Employment Equations”, The Review of Economic Studies, Vol. 58, pp.
277–297.
ARELLANO, M. AND BOVER, O. (1995): “Another look at the instrumental variable estimation of error-components models”, Journal of
Econometrics, Vol. 68, pp. 29–51.
BARBA NAVARETTI, G.; TURRINI, A. AND CHECCHI, D. (2003): “Adjusting Labour Demand: Multinational versus National Firms: A
Cross-European Analysis”, Journal of the European Economic Association, April–May 1(2–3), pp. 708 –719.
BARBA NAVARETTI, G. AND VENABLES, A. J. (2004): Multinational
Firms in the World Economy, Princeton University Press.
BERGIN, P. R.; FEENSTRA, R. C. AND HANSON, G. H. (2009): “Offshoring and Volatility: Evidence from Mexico's Maquiladora Industry”, The American Economic Review, Vol. 99, pp. 1664–1671.
BERNARD, A. B. AND SJÖHOLM, F. (2003): “Foreign Owners and Plant
Survival”, NBER Working Paper Series, Working Paper No. 10039.
BERTOLA, G. AND ROGERSON, R. (1997): “Institutions and labor reallocation”, European Economic Review, Vol 41 (6), pp. 1147–1171.
BHAGWATI, J. (1996): A New Epoch?, unpublished book review, Columbia
University
BLUNDELL, R. W. AND BOND, S. R. (1998): “Initial conditions and moment restrictions in dynamic panel datamodels”, Journal of Econometrics, Vol. 87, pp. 115–143.
BOND, S. R. (2002): “Dynamic panel data models: a guide to micro data
methods and practice”, Portuguese Economic Journal, Vol. 1, pp.
141–162.
BORENSZTEIN, E.; DE GREGORIO, J. AND LEE, J.-W. (1998): “How
does foreign direct investment affect economic growth?”, Journal of
International Economics, Vol. 45(1), pp. 115–135.
BUCH, C., M. AND LIPPONER, A. (2010): “Volatile multinationals? Evidence from the labor demand of German firms”, Labour Economics,
Vol. 17, pp. 345–353.
ECB Working Paper 1704, August 2014
36
BUCH, C. AND SCHLOTTER, M. (2013): “Regional Origins of Employment Volatility: Evidence from German states”, Empirica, Vol. 40,
pp. 1–19.
CALIENDO, M. AND KOPEINIG, S. (2008): “Some practical guidance for
the implementation of propensity score matching”, Journal of Economic Surveys, Vol. 22(1), pp. 31–72.
CUNAT, A. AND MELITZ, M., J. (2012): “Volatility, Labour Market Flexibility, and the Pattern of Comparative Advantage”, Journal of the European Economic Association, Vol. 10(2), pp. 225–254.
EAMETS, R. AND MASSO, J. (2005): “The Paradox of the Baltic States:
Labour Market Flexibility but Protected Workers?”, European Journal of Industrial Relations, Vol 11(1), pp. 71–90.
FABBRI, F.; SLAUGHTER, J., E. AND HASKEL, M., J. (2003): “Does
Nationality of Ownership Matter for Labor Demands?”, Journal of
the European Economic Association, Vol. 1, pp. 698–707.
FRANCO, C. (2013): “Exports and FDI motivations: Empirical evidence
from U.S. foreign subsidiaries”, International Business Review, Vol.
22, pp. 47–62.
GAZES, S. AND NESPOROVA, A. (2004): “Labour markets in transition:
balancing flexibility and security in Central and Eastern Europe”,
Revue de l'OFCE, Vol. 91(5), pp. 23–54.
GEISHECKER, I.; RIEDL, M. AND FRIJTERS, P. (2012): “Offshoring and
job loss fears: An econometric analysis of individual perceptions”,
Labour Economics, Vol. 19(5), pp. 738–747.
GÖRG, H.; HENRY, M.; STROBL, E. AND WALSH, F. (2009): “Multinational companies, backward linkages, and labour demand elasticities”,
Canadian Journal of Economics, Vol. 42, pp. 332–348.
GÖRG, H. AND STROBL, E. (2003): ““Footloose” multinationals”, The
Manchester School, Vol. 71, pp. 1–19.
HAKKALA, K. N.; HEYMAN, F. AND SJÖHOLM, F. (2010): “Multinationals, skills, and wage elasticities”, Review of World Economics,
Vol. 146, pp. 263–280.
HAMERMESH, D., S. (1993): Labor Demand. Princeton University Press.
HIJZEN, A. AND SWAIM P. (2010): “Offshoring, labour market institutions
and the elasticity of labour demand”, European Economic Review,
Vol. 54, pp. 1016–1034.
HANSEN, L., P. (1982): “Large Sample Properties of Generalized Method of
Moments Estimators”, Econometrica, Vol. 50, pp. 1029–1054.
LEUVEN, E. AND SIANESI, B. (2003): “PSMATCH2: Stata module to
perform full Mahalanobis and propensity score matching, common
ECB Working Paper 1704, August 2014
37
support
graphing,
and
covariate
imbalance
testing”,
[http://ideas.repec.org/c/boc/bocode/s432001.html].
LEVASSEUR, S. (2010): “International outsourcing over the business cycle:
some intuition for Germany, the Czech Republic and Slovakia”, Eastern Journal of European Studies, Vol.1, pp. 165–185.
NICKELL, S. (1981): “Biases in Dynamic Models with Fixed Effects”,
Econometrica, Vol. 49, pp. 1417–1426.
OECD (2007): Offshoring and Employment: Trends and Impacts, OECD
Publishing: 2007, 220 p.
RODRIK, D. (1997): Has globalization gone too far?, Washington, DC: Institute for International Economics, 128 p.
ROODMAN, D. (2009): “How to do xtabond2: An introduction to difference
and system GMM in Stata”, The Stata Journal, Vol. 9(1), pp. 86–136.
SCHEVE, K. AND SLAUGHTER, M. J. (2004): “Economic Insecurity and
the Globalization of Production”, American Journal of Political Science, Vol. 48( 4), pp. 662–674.
SLAUGHTER, M., J. (2001): “International trade and labor–demand elasticities”, Journal of International Economics, Vol. 54, pp. 27–56.
VISSER, J. (2011): “Data Base on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts, 1960–2010
(ICTWSS)”, Amsterdam Institute for Advanced Labour Studies
(AIAS), [http://www.uva-aias.net/208].
WAGNER, J. AND WECHE GELÜBKE, J., P. (2011): “Foreign Ownership
and Firm Survival: First Evidence for Enterprises in Germany”, IZA
Discussion Paper Series, Discussion Paper No. 6207.
ECB Working Paper 1704, August 2014
38
Appendix A: Descriptive statistics
Table 1: Descriptive statistics of domestically and foreign-owned firms in
WE countries, 2001–2009
Employment
Real wages (th of EUR)
Real turnover (th of EUR)
Real capital per employee
(th of EUR)
Real labour productivity
(th of EUR)
Age of firm
No of subsidiaries
No of shareholders
Group’s employment
Share of manufacturing
Mean
210.3
1384.0
261317.7
Domestically owned
Std. Dev. No of Obs.
3597.2
958941
11659.3
958941
4830482.0
958941
Mean
431.8
5212.9
682999.7
Foreign-owned
Std. Dev.
No of Obs.
3873.9
190154
25066.6
190154
3958836.0
190154
1768.8
39547.9
958640
4158.3
78291.3
190093
6266.4
23.8
1.76
2.45
4936.8
0.267
58843.9
15.5
16.64
4.85
8151.0
0.443
958941
957385
958941
912449
958941
958941
27686.7
27.2
2.49
1.87
2754.7
0.309
228795.4
19.5
20.80
3.19
4599.2
0.462
190154
189819
190154
176160
190154
190154
Note: The following countries are covered: Belgium, Finland, France, Germany, Italy, the
Netherlands, Norway, Portugal, Spain, Sweden and the UK.
Source: authors’ own calculations from the Amadeus dataset.
Table 2: Descriptive statistics of domestically and foreign-owned firms in
CEE countries, 2001–2009
Employment
Real wages (th of EUR)
Real turnover (th of EUR)
Real capital per employee
(th of EUR)
Real labour productivity
(th of EUR)
Age of firm
No of subsidiaries
No of shareholders
Group’s employment
Share of manufacturing
Mean
161.2
9.9
8525.7
Domestically owned
Std. Dev.
No of Obs.
621.8
66526
8.8
66526
26981.0
66526
Mean
248.7
13.1
13827.2
Foreign-owned
Std. Dev. No of Obs.
883.6
37561
15.8
37561
36123.8
37561
47.3
279.2
66457
55.9
443.3
37556
122.7
16.8
0.49
2.07
4301.0
0.314
279.9
5.4
2.30
1.67
7221.3
0.464
66526
62806
66526
65752
66526
66526
199.1
15.3
0.32
1.43
2874.9
0.393
704.5
4.6
1.60
0.92
4893.5
0.488
37561
35906
37561
36834
37561
37561
Note: The following countries are covered: Bulgaria, the Czech Republic, Estonia, Poland,
Romania, Slovakia and Slovenia.
Source: authors’ own calculations from the Amadeus dataset.
ECB Working Paper 1704, August 2014
39
Appendix B: Probit model used in propensity score
matching
Table 1: Probit model used in propensity score matching, marginal effects,
manufacturing, 2005
Log(age of firm)
Log(employment)
No of subsidiaries
Log(no of shareholders)
Log(group’s employment)
Log(capital per employee)
Log(labour productivity)
Industriesa), manufacture of
(base: food):
beverages
tobacco products
textiles
wearing apparel
leather and related products
wood and of products of wood
paper and paper products
printing and reproduction of
recorded media
coke and refined petroleum products
chemicals and chemical products
basic pharmaceutical products
rubber and plastic products
other non-metallic mineral products
basic metals
fabricated metal products
computer, electronic and optical
products
electrical equipment
machinery and equipment n.e.c.
motor vehicles
other transport equipment
furniture
other manufacturing
repair and installation of machinery and equipment
Country dummies
No of obs.
Pseudo R2
Predicted Y
ECB Working Paper 1704, August 2014
Dependent: Pr(Foreign owned=1, domestically owned=0)
Turnover volatility
Employment volatility
WE
CEE
WE
CEE
−0.012***
−0.182***
−0.013***
−0.181***
0.044***
0.040***
0.045***
0.044***
−0.001***
−0.011**
−0.001***
−0.014**
−0.050***
−0.191***
−0.050***
−0.191***
−0.032***
−0.061***
−0.032***
−0.058***
−0.004***
0.028***
−0.005***
0.029***
0.016***
0.010
0.017***
0.012*
0.021
0.149**
−0.040***
−0.038***
−0.015
−0.044***
0.062***
−0.018*
0.094*
−0.039
0.119***
0.212***
0.045
0.095**
0.183***
0.014
0.018
0.151**
−0.040***
−0.041***
−0.012
−0.046***
0.063***
−0.018*
0.088*
−0.034
0.130***
0.223***
−0.000
0.098**
0.181***
0.016
0.186***
−0.093
0.169***
−0.060
0.204***
0.255***
0.109***
0.008
0.115***
0.244***
0.219***
0.140***
0.209***
0.261***
0.109***
0.008
0.133***
0.263***
0.225***
0.152***
0.036***
0.035***
0.186***
−0.022
0.119***
0.128***
0.037***
0.036***
0.191***
−0.013
0.125***
0.138***
0.128***
0.128***
0.165***
0.002
−0.048***
0.095***
0.072***
0.249***
0.092***
0.306***
−0.049
0.092**
0.131**
−0.095**
0.128***
0.128***
0.164***
0.003
−0.050***
0.096***
0.076***
0.241***
0.092***
0.312***
−0.033
0.095**
0.137***
−0.085**
Yes
47124
0.264
0.124
Yes
7486
0.233
0.379
Yes
45705
0.265
0.125
Yes
7362
0.239
0.374
40
Dependent: Pr(Foreign owned=1, domestically owned=0)
Turnover volatility
Employment volatility
WE
CEE
WE
CEE
0.190
0.402
0.191
0.399
Actual Y
Notes: See notes for Table 2.
a)
The list of NACE Rev. 2 industries can be found at:
http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-RA-07-015/EN/KS-RA-07-015EN.PDF
ECB Working Paper 1704, August 2014
41
Table 2: Probit model used in propensity score matching, marginal effects,
services, 2005
Log(age of firm)
Log(employment)
No of subsidiaries
Log(no of shareholders)
Log(group’s employment)
Log(capital per employee)
Log(labour productivity)
Industriesa), (base: Electricity, gas,
steam):
Water collection, treatment and
supply
Sewerage
Waste collection, treatment and
disposal
Remediation activities and other
waste management
Construction of buildings
Civil engineering
Specialised construction activities
Wholesale and retail trade and repair
of motor vehicles
Wholesale trade, except of motor
vehicles
Retail trade, except of motor vehicles
Land transport and transport via
pipelines
Water transport
Air transport
Warehousing and support activities
for transportation
Postal and courier activities
Accommodation
Food and beverage service activities
Publishing activities
Motion picture, video and television
programme production
Programming and broadcasting
activities
Telecommunications
Computer programming, consultancy
and related activities
Information service activities
Financial service activities, except
insurance and pension funding
Activities auxiliary to financial services and insurance activities
Real estate activities
Legal and accounting activities
ECB Working Paper 1704, August 2014
Dependent: Pr(Foreign owned=1, domestically owned=0)
Turnover volatility
Employment volatility
WE
CEE
WE
CEE
−0.013***
−0.121***
−0.014***
−0.120***
0.020***
0.022***
0.021***
0.025***
−0.000***
0.002
−0.000***
−0.000
−0.034***
−0.157***
−0.033***
−0.160***
−0.017***
−0.030***
−0.017***
−0.028***
−0.005***
−0.007***
−0.005***
−0.007**
0.011***
0.036***
0.012***
0.039***
0.031*
−0.202***
0.038*
−0.187***
−0.026
−0.049***
−0.231***
−0.031
−0.026
−0.048***
−0.218***
−0.011
−0.065***
−0.121
−0.066***
−0.099
−0.058***
−0.048***
−0.056***
−0.029***
−0.057*
−0.090***
−0.060*
0.002
−0.060***
−0.049***
−0.056***
−0.029***
−0.041
−0.072**
−0.045
0.024
0.088***
0.172***
0.088***
0.192***
−0.042***
−0.030***
0.075**
−0.044
−0.041***
−0.029***
0.095***
−0.035
0.025*
−0.027**
0.043***
0.024
−0.005
0.125***
0.028*
−0.028**
0.044***
0.049
0.032
0.146***
−0.027*
−0.004
−0.050***
0.009
0.018
0.207
−0.052
−0.033
0.149***
0.166**
−0.022
−0.002
−0.049***
0.009
0.021
0.203
−0.047
−0.017
0.166***
0.211**
−0.015
−0.022
−0.013
0.008
0.064***
0.093***
0.231***
0.217***
0.071***
0.098***
0.226***
0.247***
0.033**
0.104***
0.047
0.309***
0.035**
0.110***
0.101
0.349***
0.056***
0.163***
0.057***
0.214***
0.006
−0.027***
0.055
0.113*
0.009
−0.021**
0.063*
0.146**
42
Dependent: Pr(Foreign owned=1, domestically owned=0)
Turnover volatility
Employment
volatility
WE
CEE
WE
CEE
Activities of head offices; management consultancy activities
Architectural and engineering
activities; technical testing and
analysis
Scientific research and development
Advertising and market research
Other professional, scientific and
technical activities
Veterinary activities
Rental and leasing activities
Employment activities
Travel agency, tour operator reservation service and related activities
Security and investigation activities
Services to buildings and landscape
activities
Office administrative, office support and other business support
activities
Public administration and defence;
compulsory social security
Education
Human health activities
Residential care activities
Social work activities without
accommodation
Creative, arts and entertainment
activities
Libraries, archives, museums and
other cultural activities
Gambling and betting activities
Sports activities and amusement
and recreation activities
Activities of membership organisations
Repair of computers and personal
and household goods
Other personal service activities
Activities of households as employers of domestic personnel
0.091***
0.296***
0.094***
0.354***
0.024**
0.042
0.023**
0.067
0.089***
0.081***
−0.195***
0.277***
0.106***
0.086***
−0.179***
0.314***
0.046***
−0.068***
−0.007
0.003
0.164*
0.239***
0.198***
0.229*
0.051***
−0.074***
−0.008
0.005
0.017
−0.021*
0.028
−0.142***
0.017
−0.020
0.044
−0.135***
−0.070***
−0.172***
−0.071***
−0.157***
0.064***
0.160***
0.067***
0.180***
−0.069***
−0.054***
−0.052***
−0.076***
−0.023
−0.213***
−0.179***
−0.068***
−0.053***
−0.050***
−0.076***
0.035
−0.195***
−0.178***
−0.089***
0.214***
0.236*
−0.091***
−0.056***
−0.090
−0.056***
−0.121
−0.076***
−0.037***
−0.156***
−0.074***
−0.041***
−0.142**
−0.048***
−0.238***
−0.047***
−0.219***
−0.080***
0.005
−0.004
−0.078***
0.043
−0.033
0.014
−0.002
0.048
0.021
−0.052
−0.051
Country dummies
Yes
Yes
Yes
Yes
No of obs.
152066
14048
143462
13745
Pseudo R2
0.227
0.255
0.226
0.259
Predicted Y
0.090
0.303
0.092
0.297
Actual Y
0.146
0.340
0.148
0.335
Notes: See notes for Table 2.
a)
The list of NACE Rev. 2 industries can be found at:
http://epp.eurostat.ec.europa.eu/cache/ITY_OFFPUB/KS-RA-07-015/EN/KS-RA-07-015EN.PDF
ECB Working Paper 1704, August 2014
43
Appendix C: Labour demand equation estimates of FOEs and DOEs: country by country
Table 1: Labour demand estimates of FOEs and DOEs, manufacturing 2001–2009
L.log(empl)
L2.log(empl)
Log(rwage)
Belgium
(lag 3 5)
Finland
(lag 3 4)
France
(lag 2 .)
wage pre
0.695***
(0.110)
Germany
(lag 2 3)
0.975***
(0.241)
−0.123
(0.156)
−0.088
(0.120)
0.795***
(0.156)
0.004
(0.145)
−0.197*
(0.114)
−0.306**
(0.128)
0.876***
(0.148)
−0.098*
(0.055)
−0.301**
(0.141)
0.089
(0.090)
0.226**
(0.089)
0.241***
(0.070)
0.178*
(0.098)
−0.308
(0.256)
0.265
(0.201)
−0.193*
(0.116)
0.105
(0.246)
−0.046
(0.198)
0.059
(0.162)
0.158*
(0.094)
0.173
(0.111)
0.102
(0.181)
−0.100
(0.078)
0.101
(0.156)
0.132
(0.092)
−0.054
(0.134)
−0.135*
(0.077)
−0.034
(0.121)
yes
yes
yes
yes
yes
yes
yes
yes
L.log(rwage)
Log(rturn)
L.log(rturn)
L.FO*log(empl)
L2.FO*log(empl)
FO*log(rwage)
L.FO*log(rwage)
FO*log(rturn)
L.FO*log(rturn)
Sector dummies
Year dummies
ECB Working Paper 1704, August 2014
Italy
(lag 3 5)
wage pre
0.761***
(0.104)
Netherlands
(lag 3 4)
−0.894***
(0.043)
0.741***
(0.096)
0.745***
(0.084)
−0.530***
(0.103)
−0.235*
(0.132)
−0.020
(0.042)
−0.226*
(0.117)
0.079
(0.085)
0.140
(0.089)
yes
yes
0.708***
(0.218)
Norway
(lag 3 5)
sec as instr
0.733***
(0.073)
Portugal
(lag 3 4)
wage pre
0.860***
(0.100)
−0.308
(0.254)
−0.253***
(0.095)
−0.143*
(0.085)
0.312*
(0.162)
0.302***
(0.082)
0.174**
(0.073)
0.030
(0.222)
−0.112
(0.195)
0.077
(0.108)
0.196
(0.252)
−0.118
(0.247)
0.073
(0.104)
−0.084
(0.178)
0.100
(0.211)
yes
yes
yes
yes
Spain
(lag 3 3)
wage pre
1.015**
(0.399)
−0.114
(0.332)
0.619
(0.546)
−0.680
(0.589)
0.180
(0.133)
Sweden
(lag 3 .)
UK
(lag 3 5)
0.800***
(0.147)
0.039
(0.134)
−0.122
(0.075)
1.018***
(0.133)
−0.156
(0.115)
−0.101
(0.072)
0.168***
(0.047)
0.126**
(0.057)
−0.146
(0.198)
0.172
(0.197)
−0.026
(0.035)
−0.029
(0.143)
0.052
(0.135)
−0.026
(0.080)
−0.071
(0.087)
−0.392
(0.416)
0.321
(0.343)
−0.139
(0.393)
0.001
(0.286)
0.102
(0.201)
0.008
(0.040)
0.002
(0.067)
no
yes
yes
yes
yes
yes
yes
yes
44
# of obs.
# of groups
Min obs. gr.
Mean obs. gr.
Max obs. gr.
# of instruments
Hansen p
AR(1) test
AR(2) test
FDI in sample
Belgium
(lag 3 5)
Finland
(lag 3 4)
11123
1716
2
6.482
7
133
0.858
−2.615
−1.795
0.379
4414
806
2
5.476
7
109
0.321
−2.785
−1.500
0.247
France
(lag 2 .)
wage pre
20695
3466
3
5.971
8
186
0.677
−7.471
−1.611
0.403
Germany
(lag 2 3)
3590
900
2
3.989
7
121
0.653
−3.190
−0.598
0.368
Italy
(lag 3 5)
wage pre
38471
5986
3
6.427
8
118
0.758
−6.505
1.152
0.167
Netherlands
(lag 3 4)
2312
364
3
6.352
8
110
0.851
−3.409
0.383
0.486
Norway
(lag 3 5)
sec as instr
7112
1795
3
3.962
6
106
0.214
−8.407
−0.451
0.066
Portugal
(lag 3 4)
wage pre
1241
254
3
4.886
8
98
0.891
−2.206
0.450
0.137
Spain
(lag 3 3)
wage pre
111476
18420
2
6.052
7
75
0.942
−1.840
0.790
0.066
Sweden
(lag 3 .)
UK
(lag 3 5)
13636
2347
2
5.810
7
169
0.176
−2.892
−0.765
0.103
29736
4890
2
6.081
7
133
0.298
−5.291
0.064
0.470
Table 1 is continued on the next page.
ECB Working Paper 1704, August 2014
45
Table 1 (continued).
L.log(empl)
Bulgaria
(lag 3 4)
0.798***
(0.183)
L2.log(empl)
Log(rwage)
Log(rturn)
L.FO*log(empl)
−0.222
(0.179)
0.328**
(0.128)
0.041
(0.149)
L2.FO*log(empl)
FO*log(rwage)
FO*log(rturn)
Sector dummies
Year dummies
# of obs.
# of groups
Min obs. gr.
Mean obs. gr.
Max obs. gr.
# of instruments
Hansen p
AR(1) test
AR(2) test
FDI in sample
Notes: See notes for Table 4.
ECB Working Paper 1704, August 2014
−0.009
(0.178)
−0.005
(0.122)
yes
yes
3518
589
3
5.973
8
110
0.445
−3.852
0.520
0.320
Czech R.
(lag 2 3)
0.746***
(0.117)
0.001
(0.029)
−0.270***
(0.098)
0.184**
(0.080)
0.070
(0.125)
0.057
(0.039)
0.125
(0.104)
−0.122
(0.094)
yes
yes
4661
850
2
5.484
7
121
0.059
−5.489
−0.089
0.786
Estonia
(lag 2 4)
0.860***
(0.255)
−0.156
(0.097)
−0.285**
(0.127)
0.301*
(0.160)
−0.102
(0.231)
0.080
(0.113)
−0.044
(0.213)
0.011
(0.174)
yes
yes
1585
304
2
5.214
6
126
0.621
−2.319
−1.487
0.539
Poland
(lag 3 .)
0.852***
(0.117)
Romania
(lag 2 .)
0.891***
(0.136)
Slovakia
(lag 2 2) wage pre
0.791***
(0.110)
Slovenia
(lag 2 4) wage ex, size*year
0.776***
(0.184)
−0.221*
(0.114)
0.196*
(0.104)
−0.028
(0.120)
−0.199
(0.136)
0.218*
(0.112)
0.052
(0.177)
−0.265**
(0.122)
0.146*
(0.081)
−0.038
(0.294)
−0.346
(0.399)
0.163
(0.149)
−0.270*
(0.160)
−0.002
(0.129)
0.003
(0.096)
yes
yes
11744
1967
3
5.971
8
170
0.108
−8.533
−1.506
0.333
0.111
(0.134)
−0.091
(0.128)
yes
yes
2230
313
3
7.125
8
218
0.964
−4.403
−1.936
0.706
−0.047
(0.176)
0.042
(0.188)
yes
yes
536
85
4
6.306
8
82
0.645
−2.780
1.121
0.670
−0.039
(0.229)
0.102
(0.171)
yes
yes
5696
908
3
6.273
7
114
0.670
−3.560
−1.045
0.129
46
Table 2: Labour demand estimates of FOEs and DOEs, services 2001–2009
Belgium
(lag 2 2)
L.log(empl)
L2.log(empl)
Log(rwage)
L.log(rwage)
Log(rturn)
L.log(rturn)
L.FO* log(empl)
L2.FO*log(empl)
FO*log(rwage)
L.FO*log(rwage)
FO*log(rturn)
L.FO*log(rturn)
Sector dummies
Year dummies
# of obs.
0.724***
(0.213)
−0.005
(0.096)
−0.833***
(0.172)
0.567***
(0.185)
0.425***
(0.146)
−0.359***
(0.139)
−0.044
(0.212)
−0.183
(0.147)
0.015
(0.205)
0.003
(0.181)
0.232
(0.174)
−0.149
(0.157)
yes
yes
27125
ECB Working Paper 1704, August 2014
Finland
(lag 3 4)
wage pre,
sec as instr
0.760***
(0.089)
France
(lag 2 .)
wage ex,
size*year
0.808***
(0.037)
−0.333***
(0.114)
−0.629***
(0.072)
0.540***
(0.089)
0.710***
(0.046)
−0.593***
(0.054)
0.042
(0.034)
0.334***
(0.095)
0.058
(0.069)
0.259***
(0.095)
−0.193***
(0.075)
yes
yes
−0.003
(0.104)
0.046
(0.110)
−0.078
(0.059)
0.039
(0.064)
no
yes
56306
Germany
(lag 2 3)
wage ex,
size*year
0.791***
(0.140)
−0.096**
(0.038)
−0.144**
(0.058)
0.139**
(0.057)
−0.024
(0.122)
−0.082
(0.055)
−0.069
(0.110)
0.070
(0.093)
no
yes
9277
Italy
(lag 2 4)
Netherlands
(lag 2 .)
wage pre
0.561***
(0.107)
Norway
(lag 3 5)
wage pre,
sec as instr
0.774***
(0.066)
Portugal
(lag 3 4)
sec as
instr
0.827***
(0.090)
Spain
(lag 3 5)
wage ex,
size*year
0.772***
(0.084)
Sweden
(lag 2 2)
wage ex,
size*year
0.766***
(0.034)
0.420***
(0.087)
−0.828***
(0.055)
0.410***
(0.072)
0.461***
(0.096)
−0.213**
(0.108)
0.091
(0.127)
−0.371***
(0.082)
−0.139**
(0.061)
−0.167**
(0.081)
−0.291***
(0.111)
0.258***
(0.096)
0.284***
(0.061)
0.205**
(0.084)
0.395***
(0.071)
0.255*
(0.144)
0.054
(0.131)
−0.130*
(0.070)
−0.185
(0.136)
−0.207***
(0.058)
0.064
(0.062)
0.563***
(0.036)
−0.422***
(0.041)
0.038
(0.045)
−0.035
(0.055)
−0.023
(0.107)
0.036
(0.104)
−0.024
(0.138)
yes
yes
44915
0.202*
(0.104)
−0.032
(0.097)
−0.084
(0.070)
−0.067
(0.181)
−0.181*
(0.103)
0.012
(0.092)
0.087
(0.059)
0.054
(0.130)
yes
yes
5332
yes
yes
44634
yes
yes
2312
yes
yes
343149
−0.010
(0.026)
−0.009
(0.026)
−0.132**
(0.061)
0.137**
(0.058)
no
yes
59948
UK
(lag 3 .)
wage ex,
size*year
0.426*
(0.249)
0.122
(0.215)
−0.264
(0.183)
0.201
(0.170)
−0.020
(0.305)
−0.046
(0.247)
0.066
(0.205)
−0.036
(0.157)
yes
yes
87274
47
Belgium
(lag 2 2)
# of groups
Min obs. gr.
Mean obs. gr.
Max obs. gr.
# of instruments
Hansen p
AR(1) test
AR(2) test
FDI in sample
4369
2
6.209
7
85
0.417
−4.578
0.869
0.408
Finland
(lag 3 4)
wage pre,
sec as instr
2512
3
6.024
8
148
0.023
−7.772
−2.876
0.264
France
(lag 2 .)
wage ex,
size*year
9709
3
5.799
8
174
0.306
−14.595
1.471
0.285
Germany
(lag 2 3)
wage ex,
size*year
2302
2
4.030
7
106
0.273
−5.604
0.419
0.223
Italy
(lag 2 4)
Netherlands
(lag 2 .)
wage pre
7651
3
5.870
8
158
0.034
−4.169
0.207
0.154
885
3
6.025
8
186
0.236
−3.937
−0.420
0.334
Norway
(lag 3 5)
wage pre,
sec as instr
11331
3
3.939
6
137
0.377
−12.187
1.147
0.071
Portugal
(lag 3 4)
sec as
instr
483
3
4.787
8
149
0.440
−5.497
−1.454
0.126
Spain
(lag 3 5)
wage ex,
size*year
50756
3
6.761
8
94
0.089
−9.265
0.070
0.058
Sweden
(lag 2 2)
wage ex,
size*year
8908
3
6.730
8
90
0.011
−27.612
1.842
0.124
UK
(lag 3 .)
wage ex,
size*year
15066
2
5.793
7
138
0.637
−1.706
−1.301
0.372
Table 2 is continued on the next page.
ECB Working Paper 1704, August 2014
48
Table 2 (continued).
L.log(empl)
Bulgaria
(lag 2 4) wage pre
0.860***
(0.099)
Czech R.
(lag 2 2)
0.671
(0.484)
Estonia
(lag 3 4) wage pre
0.840*
(0.438)
−0.224*
(0.126)
−0.741***
(0.164)
0.402
(0.393)
0.390**
(0.175)
−0.241
(0.178)
−0.127
(0.189)
−0.282
(0.228)
0.015
(0.222)
0.633***
(0.158)
−0.418
(0.260)
−0.161
(0.473)
0.007
(0.200)
0.142
(0.276)
0.097
(0.263)
−0.146
(0.157)
yes
yes
8009
1359
3
5.893
0.043
(0.251)
−0.135
(0.231)
−0.478**
(0.204)
0.635*
(0.349)
yes
yes
5678
920
3
6.172
L2.log(empl)
Log(rwage)
L.log(rwage)
Log(rturn)
0.178**
(0.081)
L.log(rturn)
L.FO* log(empl)
−0.358**
(0.153)
L2.FO*log(empl)
FO*log(rwage)
−0.012
(0.163)
L.FO*log(rwage)
FO*log(rturn)
0.123
(0.095)
L.FO*log(rturn)
Sector dummies
Year dummies
# of obs.
# of groups
Min obs. gr.
Mean obs. gr.
ECB Working Paper 1704, August 2014
yes
yes
6280
1020
3
6.157
Poland
(lag 2 .) wage pre
0.559***
(0.124)
0.089***
(0.034)
−0.656***
(0.085)
0.276***
(0.094)
0.386***
(0.071)
−0.070
(0.087)
0.084
(0.128)
−0.034
(0.045)
0.246**
(0.097)
−0.134
(0.095)
−0.216**
(0.093)
0.107
(0.105)
yes
yes
23681
4744
2
4.992
Romania
(lag 3 5) wage pre
0.610*
(0.324)
Slovakia
(lag 2 3)
0.665***
(0.141)
Slovenia
(lag 3 .) wage ex
0.719***
(0.179)
−0.258
(0.195)
−0.261*
(0.136)
−0.219
(0.211)
0.246
(0.206)
0.230*
(0.120)
0.310***
(0.101)
0.115
(0.325)
−0.018
(0.179)
−0.390
(0.318)
0.085
(0.242)
−0.107
(0.164)
0.013
(0.346)
0.041
(0.209)
0.061
(0.118)
0.102
(0.248)
yes
yes
2732
397
3
6.882
yes
yes
949
178
3
5.331
yes
yes
6815
1139
3
5.983
49
Bulgaria
(lag 2 4) wage pre
Max obs. gr.
8
# of instruments
142
Hansen p
0.005
AR(1) test
−6.645
AR(2) test
1.084
FDI in sample
0.276
Notes: See notes for Table 4.
ECB Working Paper 1704, August 2014
Czech R.
(lag 2 2)
8
92
0.257
−1.542
−0.095
0.755
Estonia
(lag 3 4) wage pre
7
79
0.042
−2.631
−0.683
0.433
Poland
(lag 2 .) wage pre
7
183
0.012
−2.533
−1.775
0.213
Romania
(lag 3 5) wage pre
8
118
0.256
−2.946
−0.503
0.787
Slovakia
(lag 2 3)
8
120
0.333
−4.836
−0.428
0.427
Slovenia
(lag 3 .) wage ex
7
89
0.016
−3.519
−2.529
0.347
50
Appendix D: Estimated speed of adjustment and
long-run elasticities: country by country
0.6
Panel (a): speed of adjustment
0.5
Domestic
firms
0.4
0.3
0.2
0.1
0
BE FI FR* DE IT* NL NO PT ES SE UK BG CZ EE PL RO SK SI*
2.5
Panel (b): absolute value of long-run wage elasticity
Domestic
firms
2
1.5
1
0.5
0
BE FI FR DE IT NL NO PT ES SE UK BG CZ EE PL RO SK SI
2.5
Panel (c): long-run output elasticity
Domestic
firms
2
1.5
1
0.5
0
BE FI FR DE IT NL NO PT ES SE UK BG CZ EE PL RO SK SI
Figure 1: Manufacturing firms’ speed of adjustment, long-run wage and
output elasticities.
Note: Based on coefficients presented in Appendix C Table 1. * indicates statistically significant difference between domestic and foreign firms at the 10% level of significance; statistical significance of difference in long-run elasticities is based on non-linear Wald-type test
using testnl command in Stata.
ECB Working Paper 1704, August 2014
51
0.7
Panel (a): speed of adjustment
0.6
Domestic
firms
0.5
0.4
0.3
0.2
0.1
0
BE FI FR DE IT NL* NO PT* ES SE UK BG* CZ EE PL RO SK SI
2
Panel (b): long-run wage elasticity
1.5
Domestic
firms
1
0.5
0
BE FI* FR DE IT* NL NO PT ES SE UK BG CZ EE PL RO SK SI
2
Panel (c): long-run output elasticity
1.5
Domestic
firms
1
0.5
0
BE FI* FR DE IT NL NO PT ES* SE UK BG CZ EE PL RO SK SI
Figure 2: Services firms’ speed of adjustment, long-run wage and output elasticities.
Note: Based on coefficients presented in Appendix C Table 2. * indicates statistically significant difference between domestic and foreign firms at the 10% level of significance; statistical significance of difference in long-run elasticities is based on non-linear Wald-type test
using testnl command in Stata.
ECB Working Paper 1704, August 2014
52
Appendix E: Labour market institutions in host and
home countries, average for 2001–2009
Average EPL
Average EPL of
home countries
of foreign firms
Average union
density
Average union
density of home
countries of
foreign firms
2.233
2.500
1.900
2.105
2.889
2.411
2.954
2.376
2.239
2.697
3.387
3.025
2.433
1.094
2.000
1.990
2.290
1.676
2.500
2.800
2.061
2.676
1.874
2.570
2.177
1.964
1.915
1.991
1.771
1.741
2.025
1.950
1.678
1.927
2.438
2.062
1.843
1.635
2.283
2.084
2.211
2.127
2.141
2.113
2.145
2.182
1.950
2.150
0.343
0.522
0.738
0.719
0.078
0.222
0.247
0.338
0.202
0.542
0.212
0.155
0.745
0.286
0.238
0.205
0.099
0.179
0.185
0.131
0.198
0.355
0.243
0.345
0.285
0.221
0.342
0.424
0.279
0.248
0.250
0.225
0.245
0.446
0.200
0.233
0.391
0.219
0.282
0.250
0.520
0.250
0.388
0.399
0.271
0.271
0.246
0.267
Sample countries
Austria
Belgium
Denmark
Finland
France
Germany
Greece
Italy
Netherlands
Norway
Portugal
Spain
Sweden
UK
Bulgaria
Czech Rep.
Estonia
Hungary
Latvia
Lithuania
Poland
Romania
Slovakia
Slovenia
Home countries of FDI in sample countries
All countries
1.879
0.261
USA
0.650
0.121
Sources: Amadeus data, ICTWSS database by Visser (2011), OECD StatExtracts.
ECB Working Paper 1704, August 2014
53
Appendix F: Monthly average labour cost, wages and
salaries (including apprentices), 2008
European Union (27 countries)
European Union (15 countries)
CEE10 average
Austria
Belgium
Denmark
Finland
Wage cost in Euros, per employee in fulltime units
3 141
3 682
1046
3 847
4 195
4 539
3 712
France
4 110
Germany
Greece
Italy
Netherlands
Norway
Portugal
Spain
Sweden
United Kingdom
Bulgaria
Czech Republic
Estonia
Latvia
Lithuania
Hungary
Poland
Romania
Slovakia
Slovenia
Note: 10 employees or more.
Source: Eurostat, LCS 2008 [lc_n08costot_r2]
3 846
2 391
3 430
4 203
5 918
1 742
2 808
4 428
3 677
374
1 323
1 149
886
848
1 164
1 089
648
991
1 991
ECB Working Paper 1704, August 2014
54